Micro_Activity_Masking_Classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "1jPihmLbUvoa"
   },
   "source": [
    "# Train Micro Activity Masking Classifiers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NWzF5-ugUvoc"
   },
   "outputs": [],
   "source": [
    "import glob\n",
    "import os\n",
    "import numpy as np\n",
    "import csv\n",
    "import io\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import shutil\n",
    "\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"] = \"PCI_BUS_ID\"\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"]=\"1\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Options to set for GPU machine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from keras.backend.tensorflow_backend import set_session\n",
    "# import tensorflow as tf\n",
    "# config = tf.ConfigProto()\n",
    "# config.gpu_options.allow_growth = True  # dynamically grow the memory used on the GPU\n",
    "# config.log_device_placement = True  # to log device placement (on which device the operation ran)\n",
    "# sess = tf.Session(config=config)\n",
    "# set_session(sess)  # set this TensorFlow session as the default"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q87cb-fJUvoo"
   },
   "outputs": [],
   "source": [
    "sub_dirs = ['left_hip','right_arm','right_wrist'] #three sensors, 9 channels total"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ho3e4qQQUvor"
   },
   "outputs": [],
   "source": [
    "#correcting the buckets information\n",
    "\n",
    "def parse_IMU_files(parent_dir, sub_dirs, startTime, endTime, file_name, window_length):\n",
    "    \n",
    "    data = []\n",
    "    data_mask = []\n",
    "    \n",
    "    data_count = 0\n",
    "    \n",
    "    for sub_dir in sub_dirs:\n",
    "        \n",
    "        channel=[[0.0, 0.0, 0.0]]*window_length\n",
    "        \n",
    "        mask = [0.0]*window_length\n",
    "        \n",
    "        div_bucket = int(10000/window_length)\n",
    "        \n",
    "        for fn in glob.glob(os.path.join(parent_dir,sub_dir, file_name)):\n",
    "            file = open(fn, newline='')\n",
    "            reader = csv.reader(file)\n",
    "            first = True\n",
    "            count = 0\n",
    "            for row in reader:\n",
    "                \n",
    "                if first:\n",
    "                    first = False\n",
    "                    continue\n",
    "                \n",
    "                timestamp=float(row[3]) #4th column is timestamp\n",
    "                if timestamp >=startTime and timestamp <=endTime and count<window_length:\n",
    "                    \n",
    "                    curr_bucket = timestamp - startTime #can go from 0 to 10,000\n",
    "                    \n",
    "                    curr_bucket = int(curr_bucket/div_bucket) #can go from 0 to 500\n",
    "                    \n",
    "                    #print('curr_bucket :',curr_bucket)\n",
    "                    #channel.append([float(row[0]),float(row[1]),float(row[2])])\n",
    "                    \n",
    "                    if curr_bucket<window_length:\n",
    "                        channel[curr_bucket] = [float(row[0]),float(row[1]),float(row[2])]\n",
    "                        mask[curr_bucket] = 1.0\n",
    "\n",
    "                        count = count + 1 \n",
    "                        data_count = data_count+1\n",
    "                    \n",
    "                    \n",
    "        data.append(channel) \n",
    "        data_mask.append(mask)\n",
    "    return data, data_mask, data_count\n",
    "\n",
    "\n",
    "\n",
    "#correcting the buckets information\n",
    "\n",
    "import random\n",
    "\n",
    "def parse_IMU_files_2(parent_dir, sub_dirs, startTime, endTime, file_name, window_length):\n",
    "    \n",
    "    data = []\n",
    "    data_mask = []\n",
    "    \n",
    "    data_count = 0\n",
    "    \n",
    "    for sub_dir in sub_dirs:\n",
    "        \n",
    "        channel=[[0.0, 0.0, 0.0]]*window_length\n",
    "        \n",
    "        mask = [0.0]*window_length\n",
    "        \n",
    "        div_bucket = int(10000/window_length)\n",
    "        \n",
    "        for fn in glob.glob(os.path.join(parent_dir,sub_dir, file_name)):\n",
    "            file = open(fn, newline='')\n",
    "            reader = csv.reader(file)\n",
    "            first = True\n",
    "            count = 0\n",
    "            for row in reader:\n",
    "                \n",
    "                if first:\n",
    "                    first = False\n",
    "                    continue\n",
    "                \n",
    "                timestamp=float(row[3]) #4th column is timestamp\n",
    "                \n",
    "                window_jitter1 = random.randint(-150,150)\n",
    "                \n",
    "                window_jitter2 = random.randint(-150,150)\n",
    "                \n",
    "                \n",
    "                if timestamp >=(startTime+window_jitter1) and timestamp <=(endTime+window_jitter2) and count<window_length:\n",
    "                    \n",
    "                    curr_bucket = timestamp - startTime #can go from 0 to 10,000\n",
    "                    \n",
    "                    curr_bucket = int(curr_bucket/div_bucket) #can go from 0 to 500\n",
    "                    \n",
    "                    if curr_bucket<window_length:\n",
    "                        channel[curr_bucket] = [float(row[0]),float(row[1]),float(row[2])]\n",
    "                        mask[curr_bucket] = 1.0\n",
    "\n",
    "                        count = count + 1 \n",
    "                        data_count = data_count+1                   \n",
    "                    \n",
    "        data.append(channel) \n",
    "        data_mask.append(mask)\n",
    "    return data, data_mask, data_count\n",
    "\n",
    "\n",
    "\n",
    "def parse_IMU_files_3(parent_dir, sub_dirs, startTime, endTime, file_name, window_length):\n",
    "    \n",
    "    data = []\n",
    "    data_mask = []\n",
    "    \n",
    "    data_count = 0\n",
    "    \n",
    "    for sub_dir in sub_dirs:\n",
    "        \n",
    "        channel=[[0.0, 0.0, 0.0]]*window_length\n",
    "        \n",
    "        mask = [0.0]*window_length\n",
    "        \n",
    "        div_bucket = int(10000/window_length)\n",
    "        \n",
    "        window_jitter1 = random.randint(-1500,1500)\n",
    "        window_jitter2 = random.randint(-1500,1500)\n",
    "                \n",
    "        \n",
    "        for fn in glob.glob(os.path.join(parent_dir,sub_dir, file_name)):\n",
    "            file = open(fn, newline='')\n",
    "            reader = csv.reader(file)\n",
    "            first = True\n",
    "            count = 0\n",
    "            for row in reader:\n",
    "                \n",
    "                if first:\n",
    "                    first = False\n",
    "                    continue\n",
    "                \n",
    "                timestamp=float(row[3]) #4th column is timestamp\n",
    "                \n",
    "                \n",
    "                if timestamp >=(startTime+window_jitter1) and timestamp <=(endTime+window_jitter2) and count<window_length:\n",
    "                    \n",
    "                    curr_bucket = timestamp - startTime #can go from 0 to 10,000\n",
    "                    \n",
    "                    curr_bucket = int(curr_bucket/div_bucket) #can go from 0 to 500\n",
    "                    \n",
    "                    if curr_bucket<window_length:\n",
    "                        channel[curr_bucket] = [float(row[0]),float(row[1]),float(row[2])]\n",
    "                        mask[curr_bucket] = 1.0\n",
    "\n",
    "                        count = count + 1 \n",
    "                        data_count = data_count+1                   \n",
    "                    \n",
    "        data.append(channel) \n",
    "        data_mask.append(mask)\n",
    "    return data, data_mask, data_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_min_count = 100 # per sample, valid points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DLINn0siUvou"
   },
   "outputs": [],
   "source": [
    "def get_train_data(data_dir, sub_dirs, pos_lab, neg_lab, labeltable):\n",
    "    files = os.listdir(data_dir+'/left_hip')\n",
    "    number_of_samples = 500\n",
    "    \n",
    "    labels_macro =dict()\n",
    "    \n",
    "    labels_macro[pos_lab] = 0\n",
    "    labels_macro[neg_lab] = 1\n",
    "    \n",
    "    \n",
    "    #read the labels\n",
    "    labels_loc = labeltable\n",
    "    file_label = open(labels_loc, newline='')\n",
    "    label_reader = csv.reader(file_label)\n",
    "    file_label_mapping = dict()\n",
    "    \n",
    "    first = True\n",
    "    for row in label_reader:\n",
    "        \n",
    "        if first:\n",
    "            first = False\n",
    "            continue\n",
    "        \n",
    "        file_label_mapping[row[0]+'.csv'] = labels_macro[row[2]]\n",
    "        \n",
    "    all_data = []\n",
    "    all_labels = []\n",
    "    all_filename = []\n",
    "    for f in files:\n",
    "        \n",
    "        st_index = 0\n",
    "        end_index = 30000\n",
    "        step = 1000 #overlapping window, step: 1000. \n",
    "        window_index = 10000 #6 second window\n",
    "        \n",
    "        #print('reading file:',f)\n",
    "        f_name = f\n",
    "        \n",
    "        if f_name == '.DS_Store':\n",
    "            continue\n",
    "        \n",
    "        curr_label_file = file_label_mapping[f_name]\n",
    "        \n",
    "        while st_index+step < end_index:\n",
    "        \n",
    "            data, data_mask, data_count = parse_IMU_files(data_dir, sub_dirs, st_index, st_index+window_index,  f, number_of_samples)\n",
    "            st_index = st_index+step\n",
    "            \n",
    "            if data_count<data_min_count:\n",
    "                continue\n",
    "            \n",
    "            train_data_sample  = np.zeros((12, number_of_samples))\n",
    "            train_data_label   = curr_label_file\n",
    "            for i in range(len(data)):\n",
    "                for j in range(len(data[i])):\n",
    "                    train_data_sample[i*4,j]=data[i][j][0]\n",
    "                    train_data_sample[i*4+1,j]=data[i][j][1]\n",
    "                    train_data_sample[i*4+2,j]=data[i][j][2]\n",
    "                    \n",
    "                    #mask information\n",
    "                    train_data_sample[i*4+3,j]=data_mask[i][j]\n",
    "                    \n",
    "                    \n",
    "            all_filename.append(f_name)\n",
    "            all_data.append(train_data_sample)\n",
    "            all_labels.append(train_data_label)\n",
    "            \n",
    "            \n",
    "            \n",
    "    return all_data, all_labels, all_filename\n",
    "\n",
    "\n",
    "def get_train_data_2(data_dir, sub_dirs,pos_lab, neg_lab, labeltable):\n",
    "    files = os.listdir(data_dir+'/left_hip')\n",
    "    number_of_samples = 500\n",
    "    \n",
    "    labels_macro =dict()\n",
    "    \n",
    "    labels_macro[pos_lab] = 0\n",
    "    labels_macro[neg_lab] = 1\n",
    "    \n",
    "    \n",
    "    #read the labels\n",
    "    labels_loc = labeltable\n",
    "    file_label = open(labels_loc, newline='')\n",
    "    label_reader = csv.reader(file_label)\n",
    "    file_label_mapping = dict()\n",
    "    \n",
    "    first = True\n",
    "    for row in label_reader:\n",
    "        \n",
    "        if first:\n",
    "            first = False\n",
    "            continue\n",
    "        \n",
    "        file_label_mapping[row[0]+'.csv'] = labels_macro[row[2]]\n",
    "        \n",
    "    all_data = []\n",
    "    all_labels = []\n",
    "    \n",
    "    all_counts = []\n",
    "    \n",
    "    for f in files:\n",
    "        \n",
    "        st_index = 0\n",
    "        end_index = 30000\n",
    "        step = 1000 #overlapping window, step \n",
    "        window_index = 10000 #6 second window\n",
    "        \n",
    "        #print('reading file:',f)\n",
    "        f_name = f\n",
    "        \n",
    "        if f_name == '.DS_Store':\n",
    "            continue\n",
    "        \n",
    "        curr_label_file = file_label_mapping[f_name]\n",
    "        \n",
    "        while st_index+step < end_index:\n",
    "        \n",
    "            data, data_mask, data_count = parse_IMU_files_2(data_dir, sub_dirs, st_index, st_index+window_index,  f, number_of_samples)\n",
    "            \n",
    "            st_index = st_index+step\n",
    "            \n",
    "            if data_count<data_min_count:\n",
    "                continue\n",
    "            \n",
    "            train_data_sample  = np.zeros((12, number_of_samples))\n",
    "            train_data_label   = curr_label_file\n",
    "            for i in range(len(data)):\n",
    "                for j in range(len(data[i])):\n",
    "                    train_data_sample[i*4,j]=data[i][j][0]\n",
    "                    train_data_sample[i*4+1,j]=data[i][j][1]\n",
    "                    train_data_sample[i*4+2,j]=data[i][j][2]\n",
    "                    \n",
    "                    #mask information\n",
    "                    train_data_sample[i*4+3,j]=data_mask[i][j]\n",
    "            \n",
    "            all_data.append(train_data_sample)\n",
    "            all_labels.append(train_data_label)\n",
    "            \n",
    "            all_counts.append(data_count)\n",
    "            \n",
    "            \n",
    "            \n",
    "    return all_data, all_labels, all_counts\n",
    "\n",
    "\n",
    "def get_train_data_3(data_dir, sub_dirs,pos_lab, neg_lab, labeltable):\n",
    "    files = os.listdir(data_dir+'/left_hip')\n",
    "    number_of_samples = 500\n",
    "    \n",
    "    labels_macro =dict()\n",
    "    \n",
    "    labels_macro[pos_lab] = 0\n",
    "    labels_macro[neg_lab] = 1\n",
    "    \n",
    "    \n",
    "    #read the labels\n",
    "    labels_loc = labeltable\n",
    "    file_label = open(labels_loc, newline='')\n",
    "    label_reader = csv.reader(file_label)\n",
    "    file_label_mapping = dict()\n",
    "    \n",
    "    first = True\n",
    "    for row in label_reader:\n",
    "        \n",
    "        if first:\n",
    "            first = False\n",
    "            continue\n",
    "        \n",
    "        file_label_mapping[row[0]+'.csv'] = labels_macro[row[2]]\n",
    "        \n",
    "    all_data = []\n",
    "    all_labels = []\n",
    "    \n",
    "    all_counts = []\n",
    "    for f in files:\n",
    "        \n",
    "        st_index = 0\n",
    "        end_index = 30000\n",
    "        step = 1000 #overlapping window, step: 1000. \n",
    "        window_index = 10000 #6 second window\n",
    "        \n",
    "        #print('reading file:',f)\n",
    "        f_name = f\n",
    "        \n",
    "        if f_name == '.DS_Store':\n",
    "            continue\n",
    "        \n",
    "        curr_label_file = file_label_mapping[f_name]\n",
    "        \n",
    "        while st_index+step < end_index:\n",
    "                \n",
    "            \n",
    "            data, data_mask, data_count  = parse_IMU_files_3(data_dir, sub_dirs, st_index, st_index+window_index,  f, number_of_samples)\n",
    "            \n",
    "            st_index = st_index+step\n",
    "            \n",
    "            if data_count<data_min_count:\n",
    "                continue\n",
    "            \n",
    "            train_data_sample  = np.zeros((12, number_of_samples))\n",
    "            train_data_label   = curr_label_file\n",
    "            for i in range(len(data)):\n",
    "                for j in range(len(data[i])):\n",
    "                    train_data_sample[i*4,j]=data[i][j][0]\n",
    "                    train_data_sample[i*4+1,j]=data[i][j][1]\n",
    "                    train_data_sample[i*4+2,j]=data[i][j][2]\n",
    "                    \n",
    "                    #mask information\n",
    "                    train_data_sample[i*4+3,j]=data_mask[i][j]\n",
    "                    \n",
    "\n",
    "            all_data.append(train_data_sample)\n",
    "            all_labels.append(train_data_label)\n",
    "            \n",
    "            all_counts.append(data_count)\n",
    "            \n",
    "            \n",
    "    return all_data, all_labels, all_counts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_prediction_accuracy(data_dir, sub_dirs,pos_lab,neg_lab,labeltable,model):\n",
    "\n",
    "    files = os.listdir(data_dir+'/left_hip')\n",
    "    number_of_samples = 500\n",
    "\n",
    "    labels_macro =dict()\n",
    "\n",
    "    labels_macro[pos_lab] = 0\n",
    "    labels_macro[neg_lab] = 1\n",
    "\n",
    "\n",
    "    #read the labels\n",
    "    labels_loc = labeltable\n",
    "    file_label = open(labels_loc, newline='')\n",
    "    label_reader = csv.reader(file_label)\n",
    "    file_label_mapping = dict()\n",
    "\n",
    "    first = True\n",
    "    for row in label_reader:\n",
    "\n",
    "        if first:\n",
    "            first = False\n",
    "            continue\n",
    "\n",
    "        file_label_mapping[row[0]+'.csv'] = labels_macro[row[2]]\n",
    "\n",
    "\n",
    "    total_count = 0\n",
    "    correct_count = 0\n",
    "\n",
    "    for f in files:\n",
    "\n",
    "        file_data = []\n",
    "        file_label = []\n",
    "\n",
    "        st_index = 0\n",
    "        end_index = 30000\n",
    "        step = 1000 #overlapping window, step: 1000. \n",
    "        window_index = 10000 #6 second window\n",
    "\n",
    "        #print('reading file:',f)\n",
    "        f_name = f\n",
    "\n",
    "        if f_name == '.DS_Store':\n",
    "            continue\n",
    "\n",
    "        total_count = total_count+1\n",
    "\n",
    "        curr_label_file = file_label_mapping[f_name]\n",
    "\n",
    "        while st_index+step < end_index:\n",
    "\n",
    "            data, data_mask, data_count = parse_IMU_files(data_dir, sub_dirs, st_index, st_index+window_index,  f, number_of_samples)\n",
    "            st_index = st_index+step\n",
    "\n",
    "            if data_count<data_min_count:\n",
    "                continue\n",
    "\n",
    "            train_data_sample  = np.zeros((12, number_of_samples))\n",
    "            train_data_label   = curr_label_file\n",
    "            for i in range(len(data)):\n",
    "                for j in range(len(data[i])):\n",
    "                    train_data_sample[i*4,j]=data[i][j][0]\n",
    "                    train_data_sample[i*4+1,j]=data[i][j][1]\n",
    "                    train_data_sample[i*4+2,j]=data[i][j][2]\n",
    "                    \n",
    "                    #mask information\n",
    "                    train_data_sample[i*4+3,j]=data_mask[i][j]\n",
    "\n",
    "            file_data.append(train_data_sample)\n",
    "            #file_label.append(train_data_label)\n",
    "\n",
    "        file_data = np.array(file_data)\n",
    "        file_label = curr_label_file\n",
    "\n",
    "        file_data = file_data.reshape((-1, 12,500, 1))\n",
    "\n",
    "        y_pred = np.argmax(model.predict(file_data), axis=1) \n",
    "        counts = np.bincount(y_pred)\n",
    "        prediction = np.argmax(counts) #max occuring value in windows\n",
    "        \n",
    "        #correct prediction\n",
    "        if int(prediction)==int(file_label):\n",
    "            correct_count = correct_count+1\n",
    "        \n",
    "    \n",
    "    return total_count, correct_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, LSTM, Dense, Dropout, Flatten, Bidirectional\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.core import Permute, Reshape\n",
    "from keras import backend as K\n",
    "from keras.callbacks import ModelCheckpoint"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def _data_reshaping(X_tr, X_va, X_ts, network_type):\n",
    "    _, win_len, dim = X_tr.shape\n",
    "    print(network_type)\n",
    "    if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "\n",
    "        X_tr = np.swapaxes(X_tr,1,2)\n",
    "        X_va = np.swapaxes(X_va,1,2)\n",
    "        X_ts = np.swapaxes(X_ts,1,2)\n",
    "\n",
    "        X_tr = np.reshape(X_tr, (-1, dim, win_len, 1))\n",
    "        X_va = np.reshape(X_va, (-1, dim, win_len, 1))\n",
    "        X_ts = np.reshape(X_ts, (-1, dim, win_len, 1))\n",
    "    \n",
    "    return X_tr, X_va, X_ts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import regularizers\n",
    "\n",
    "def model_variant(model, num_feat_map, dim, network_type,p):\n",
    "    print(network_type)\n",
    "    if network_type == 'ConvLSTM':\n",
    "        model.add(Permute((2, 1, 3))) \n",
    "        model.add(Reshape((-1,num_feat_map*dim)))\n",
    "        model.add(Bidirectional(LSTM(128, return_sequences=False, stateful=False)))\n",
    "    if network_type == 'CNN':\n",
    "        \n",
    "        model.add(Flatten())\n",
    "        model.add(Dense(64, activation='relu'))\n",
    "        model.add(BatchNormalization()) \n",
    "        model.add(Dropout(p))\n",
    "\n",
    "        \n",
    "def model_conv(model, num_feat_map,p,b):\n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10),    \n",
    "                 activation='relu',\n",
    "                 input_shape=(dim, win_len, 1),\n",
    "                 padding='same'))\n",
    "    \n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10), activation='relu',padding='same'))\n",
    "    \n",
    "    if (b==1):\n",
    "        model.add(BatchNormalization()) \n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10), activation='relu',padding='same'))\n",
    "    \n",
    "    if (b==1):\n",
    "        model.add(BatchNormalization()) \n",
    "    model.add(MaxPooling2D(pool_size=(1, 3)))\n",
    "    \n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10), activation='relu',padding='same')) \n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10), activation='relu',padding='same'))\n",
    "    if (b==1):\n",
    "        model.add(BatchNormalization()) \n",
    "    model.add(MaxPooling2D(pool_size=(1, 2)))\n",
    "    model.add(Dropout(p))\n",
    "    \n",
    "    model.add(Conv2D(num_feat_map, kernel_size=(1, 10), activation='relu',padding='same'))  \n",
    "    if (b==1):\n",
    "        model.add(BatchNormalization()) \n",
    "    model.add(MaxPooling2D(pool_size=(1, 2)))\n",
    "    \n",
    "    model.add(Dropout(p))\n",
    "    \n",
    "def model_LSTM(model,p):\n",
    "    model.add(LSTM(num_hidden_lstm, \n",
    "               input_shape=(win_len,dim), \n",
    "               return_sequences=True))\n",
    "    model.add(Dropout(p))\n",
    "    model.add(LSTM(num_hidden_lstm, return_sequences=False))\n",
    "    model.add(Dropout(p))\n",
    "    \n",
    "def model_output(model):\n",
    "    model.add(Dense(num_classes, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import f1_score, accuracy_score\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "\n",
    "class monitor_Training(keras.callbacks.Callback):\n",
    "    def on_train_begin(self, logs={}):\n",
    "        return\n",
    " \n",
    "    def on_train_end(self, logs={}):\n",
    "        return\n",
    " \n",
    "    def on_epoch_begin(self, epoch, logs={}):\n",
    "        return\n",
    " \n",
    "    def on_epoch_end(self, epoch, logs={}):\n",
    "        y_pred = np.argmax(self.model.predict(X_test), axis=1)\n",
    "        accuracy = accuracy_score(y_true, y_pred)\n",
    "        print('test accuracy is:', accuracy)\n",
    "        return\n",
    " \n",
    "    def on_batch_begin(self, batch, logs={}):\n",
    "        return\n",
    " \n",
    "    def on_batch_end(self, batch, logs={}):\n",
    "        return"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training/Testing:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Add"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_add_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_add_TTV/val'\n",
    "data_dir_test='data_micro/data_m_add_TTV/test'\n",
    "pos_lab = 'Add'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelAdd.csv'\n",
    "model_name = \"data_micro/DCBL_micro_ADD_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5053, 12, 500)\n",
      "(5053,)\n",
      "(5052, 12, 500)\n",
      "(5052,)\n",
      "(5054, 12, 500)\n",
      "(5054,)\n",
      "(1382, 12, 500)\n",
      "(1382,)\n",
      "(1378, 12, 500)\n",
      "(1378,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15159, 12, 500)\n",
      "(15159,)\n",
      "(1382, 12, 500)\n",
      "(1382,)\n",
      "(1378, 12, 500)\n",
      "(1378,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15159, 12, 500)\n",
      "(15159, 2)\n",
      "(1382, 12, 500)\n",
      "(1382, 2)\n",
      "(1378, 12, 500)\n",
      "(1378,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "WARNING:tensorflow:From /home/sandeep/anaconda3/envs/sandeep-env/lib/python3.7/site-packages/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "If using Keras pass *_constraint arguments to layers.\n",
      "WARNING:tensorflow:From /home/sandeep/anaconda3/envs/sandeep-env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4070: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
      "\n",
      "ConvLSTM\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_1 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_1 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_1 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15159, 12, 500, 1) (1382, 12, 500, 1) (1378, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.9580318331718445\n",
      "Training accuracy of best model:  0.9327792\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[   0   84]\n",
      " [   0 1294]]\n",
      "the mean-f1 score: 0.48\n",
      "accuracy is: 0.94\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 47\n",
      "Percentage accuracy: 94.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 175\n",
      "Percentage accuracy: 93.08510638297872\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 48\n",
      "Percentage accuracy: 96.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Cut"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_cut_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_cut_TTV/val'\n",
    "data_dir_test='data_micro/data_m_cut_TTV/test'\n",
    "pos_lab = 'Cut'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelCut.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Cut_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5073, 12, 500)\n",
      "(5073,)\n",
      "(5060, 12, 500)\n",
      "(5060,)\n",
      "(5071, 12, 500)\n",
      "(5071,)\n",
      "(1416, 12, 500)\n",
      "(1416,)\n",
      "(1327, 12, 500)\n",
      "(1327,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15204, 12, 500)\n",
      "(15204,)\n",
      "(1416, 12, 500)\n",
      "(1416,)\n",
      "(1327, 12, 500)\n",
      "(1327,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15204, 12, 500)\n",
      "(15204, 2)\n",
      "(1416, 12, 500)\n",
      "(1416, 2)\n",
      "(1327, 12, 500)\n",
      "(1327,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_7 (Conv2D)            (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_10 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_11 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_7 (Batch (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_12 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_8 (Batch (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_2 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_2 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_2 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15204, 12, 500, 1) (1416, 12, 500, 1) (1327, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.7853107452392578\n",
      "Training accuracy of best model:  0.97842675\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[304 131]\n",
      " [141 751]]\n",
      "the mean-f1 score: 0.77\n",
      "accuracy is: 0.80\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 44\n",
      "Percentage accuracy: 88.0\n",
      "total_count of train files: 187\n",
      "correctly predicted train files: 187\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 51\n",
      "correctly predicted val files: 44\n",
      "Percentage accuracy: 86.27450980392157\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Wash"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_wash_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_wash_TTV/val'\n",
    "data_dir_test='data_micro/data_m_wash_TTV/test'\n",
    "pos_lab = 'Wash'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelWash.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Wash_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5072, 12, 500)\n",
      "(5072,)\n",
      "(5063, 12, 500)\n",
      "(5063,)\n",
      "(5072, 12, 500)\n",
      "(5072,)\n",
      "(1411, 12, 500)\n",
      "(1411,)\n",
      "(1331, 12, 500)\n",
      "(1331,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15207, 12, 500)\n",
      "(15207,)\n",
      "(1411, 12, 500)\n",
      "(1411,)\n",
      "(1331, 12, 500)\n",
      "(1331,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15207, 12, 500)\n",
      "(15207, 2)\n",
      "(1411, 12, 500)\n",
      "(1411, 2)\n",
      "(1331, 12, 500)\n",
      "(1331,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_13 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_14 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_9 (Batch (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_15 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_10 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_16 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_17 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_11 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2 (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_18 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_12 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_9 (MaxPooling2 (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_3 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_3 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_3 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15207, 12, 500, 1) (1411, 12, 500, 1) (1331, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8901488184928894\n",
      "Training accuracy of best model:  0.98908395\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  96  136]\n",
      " [  38 1061]]\n",
      "the mean-f1 score: 0.72\n",
      "accuracy is: 0.87\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 44\n",
      "Percentage accuracy: 88.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 188\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 46\n",
      "Percentage accuracy: 92.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Take"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_take_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_take_TTV/val'\n",
    "data_dir_test='data_micro/data_m_take_TTV/test'\n",
    "pos_lab = 'Take'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelTake.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Take_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5096, 12, 500)\n",
      "(5096,)\n",
      "(5094, 12, 500)\n",
      "(5094,)\n",
      "(5096, 12, 500)\n",
      "(5096,)\n",
      "(1382, 12, 500)\n",
      "(1382,)\n",
      "(1336, 12, 500)\n",
      "(1336,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15286, 12, 500)\n",
      "(15286,)\n",
      "(1382, 12, 500)\n",
      "(1382,)\n",
      "(1336, 12, 500)\n",
      "(1336,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15286, 12, 500)\n",
      "(15286, 2)\n",
      "(1382, 12, 500)\n",
      "(1382, 2)\n",
      "(1336, 12, 500)\n",
      "(1336,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_19 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_20 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_13 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_21 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_14 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_10 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_22 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_23 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_15 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_11 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_24 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_16 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_12 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_4 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_4 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_4 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15286, 12, 500, 1) (1382, 12, 500, 1) (1336, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8248914480209351\n",
      "Training accuracy of best model:  0.9960094\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[543  62]\n",
      " [140 591]]\n",
      "the mean-f1 score: 0.84\n",
      "accuracy is: 0.85\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 45\n",
      "Percentage accuracy: 90.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 188\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 46\n",
      "Percentage accuracy: 92.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Put"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_put_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_put_TTV/val'\n",
    "data_dir_test='data_micro/data_m_put_TTV/test'\n",
    "pos_lab = 'Put'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelPut.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Put_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5116, 12, 500)\n",
      "(5116,)\n",
      "(5117, 12, 500)\n",
      "(5117,)\n",
      "(5117, 12, 500)\n",
      "(5117,)\n",
      "(1344, 12, 500)\n",
      "(1344,)\n",
      "(1353, 12, 500)\n",
      "(1353,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15350, 12, 500)\n",
      "(15350,)\n",
      "(1344, 12, 500)\n",
      "(1344,)\n",
      "(1353, 12, 500)\n",
      "(1353,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15350, 12, 500)\n",
      "(15350, 2)\n",
      "(1344, 12, 500)\n",
      "(1344, 2)\n",
      "(1353, 12, 500)\n",
      "(1353,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_25 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_26 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_17 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_27 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_18 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_13 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_28 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_29 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_19 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_14 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_9 (Dropout)          (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_30 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_20 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_15 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_10 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_5 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_5 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_5 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15350, 12, 500, 1) (1344, 12, 500, 1) (1353, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8147321343421936\n",
      "Training accuracy of best model:  0.9879479\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[341 113]\n",
      " [109 790]]\n",
      "the mean-f1 score: 0.81\n",
      "accuracy is: 0.84\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 43\n",
      "Percentage accuracy: 86.0\n",
      "total_count of train files: 189\n",
      "correctly predicted train files: 186\n",
      "Percentage accuracy: 98.4126984126984\n",
      "total_count of val files: 49\n",
      "correctly predicted val files: 45\n",
      "Percentage accuracy: 91.83673469387756\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Peel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_peel_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_peel_TTV/val'\n",
    "data_dir_test='data_micro/data_m_peel_TTV/test'\n",
    "pos_lab = 'Peel'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelPeel.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Peel_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5163, 12, 500)\n",
      "(5163,)\n",
      "(5159, 12, 500)\n",
      "(5159,)\n",
      "(5163, 12, 500)\n",
      "(5163,)\n",
      "(1280, 12, 500)\n",
      "(1280,)\n",
      "(1371, 12, 500)\n",
      "(1371,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15485, 12, 500)\n",
      "(15485,)\n",
      "(1280, 12, 500)\n",
      "(1280,)\n",
      "(1371, 12, 500)\n",
      "(1371,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15485, 12, 500)\n",
      "(15485, 2)\n",
      "(1280, 12, 500)\n",
      "(1280, 2)\n",
      "(1371, 12, 500)\n",
      "(1371,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_6\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_31 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_32 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_21 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_33 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_22 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_16 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_34 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_35 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_23 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_17 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_11 (Dropout)         (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_36 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_24 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_18 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_12 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_6 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_6 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_6 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15485, 12, 500, 1) (1280, 12, 500, 1) (1371, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8500000238418579\n",
      "Training accuracy of best model:  0.99916047\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[374  32]\n",
      " [113 852]]\n",
      "the mean-f1 score: 0.88\n",
      "accuracy is: 0.89\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 49\n",
      "Percentage accuracy: 98.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 188\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 46\n",
      "Percentage accuracy: 92.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_mix_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_mix_TTV/val'\n",
    "data_dir_test='data_micro/data_m_mix_TTV/test'\n",
    "pos_lab = 'Mix'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelMix.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Mix_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5044, 12, 500)\n",
      "(5044,)\n",
      "(5035, 12, 500)\n",
      "(5035,)\n",
      "(5044, 12, 500)\n",
      "(5044,)\n",
      "(1407, 12, 500)\n",
      "(1407,)\n",
      "(1363, 12, 500)\n",
      "(1363,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15123, 12, 500)\n",
      "(15123,)\n",
      "(1407, 12, 500)\n",
      "(1407,)\n",
      "(1363, 12, 500)\n",
      "(1363,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15123, 12, 500)\n",
      "(15123, 2)\n",
      "(1407, 12, 500)\n",
      "(1407, 2)\n",
      "(1363, 12, 500)\n",
      "(1363,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_7\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_37 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_38 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_25 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_39 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_26 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_19 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_40 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_41 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_27 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_20 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_13 (Dropout)         (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_42 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_28 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_21 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_14 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_7 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_7 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_7 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15123, 12, 500, 1) (1407, 12, 500, 1) (1363, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.9545131325721741\n",
      "Training accuracy of best model:  0.9990743\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  40   63]\n",
      " [  14 1246]]\n",
      "the mean-f1 score: 0.74\n",
      "accuracy is: 0.94\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 47\n",
      "Percentage accuracy: 94.0\n",
      "total_count of train files: 187\n",
      "correctly predicted train files: 187\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 51\n",
      "correctly predicted val files: 49\n",
      "Percentage accuracy: 96.07843137254902\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Pour"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_pour_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_pour_TTV/val'\n",
    "data_dir_test='data_micro/data_m_pour_TTV/test'\n",
    "pos_lab = 'Pour'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelPour.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Pour_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5205, 12, 500)\n",
      "(5205,)\n",
      "(5193, 12, 500)\n",
      "(5193,)\n",
      "(5205, 12, 500)\n",
      "(5205,)\n",
      "(1274, 12, 500)\n",
      "(1274,)\n",
      "(1335, 12, 500)\n",
      "(1335,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15603, 12, 500)\n",
      "(15603,)\n",
      "(1274, 12, 500)\n",
      "(1274,)\n",
      "(1335, 12, 500)\n",
      "(1335,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15603, 12, 500)\n",
      "(15603, 2)\n",
      "(1274, 12, 500)\n",
      "(1274, 2)\n",
      "(1335, 12, 500)\n",
      "(1335,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_8\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_43 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_44 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_29 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_45 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_30 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_22 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_46 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_47 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_31 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_23 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_15 (Dropout)         (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_48 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_32 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_24 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_16 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_8 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_8 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_8 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15603, 12, 500, 1) (1274, 12, 500, 1) (1335, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.9167974591255188\n",
      "Training accuracy of best model:  0.99442416\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  30  117]\n",
      " [   8 1180]]\n",
      "the mean-f1 score: 0.64\n",
      "accuracy is: 0.91\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 45\n",
      "Percentage accuracy: 90.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 185\n",
      "Percentage accuracy: 98.40425531914893\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 46\n",
      "Percentage accuracy: 92.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Open"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_open_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_open_TTV/val'\n",
    "data_dir_test='data_micro/data_m_open_TTV/test'\n",
    "pos_lab = 'Open'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'LabelOpen.csv'\n",
    "model_name = \"data_micro/DCBL_micro_Open_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5045, 12, 500)\n",
      "(5045,)\n",
      "(5040, 12, 500)\n",
      "(5040,)\n",
      "(5046, 12, 500)\n",
      "(5046,)\n",
      "(1428, 12, 500)\n",
      "(1428,)\n",
      "(1340, 12, 500)\n",
      "(1340,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15131, 12, 500)\n",
      "(15131,)\n",
      "(1428, 12, 500)\n",
      "(1428,)\n",
      "(1340, 12, 500)\n",
      "(1340,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15131, 12, 500)\n",
      "(15131, 2)\n",
      "(1428, 12, 500)\n",
      "(1428, 2)\n",
      "(1340, 12, 500)\n",
      "(1340,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_9\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_49 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_50 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_33 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_51 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_34 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_25 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_52 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_53 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_35 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_26 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_17 (Dropout)         (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_54 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_36 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_27 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_18 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_9 (Permute)          (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_9 (Reshape)          (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_9 (Bidirection (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15131, 12, 500, 1) (1428, 12, 500, 1) (1340, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEWCAYAAABrDZDcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3de5xcdX3/8ddnZu/ZzXWXJLCBRAmQhIvAiii1VsH+QBQsikKLFYtQWrmJtuLlYam1P6miVSi/8kOLiFCutooWRX5KChZUAoQNCYIhQLJhJ9mE3De7O7vz+f1xzu6e3exmJ8nMnpk57+eDecy5z2cm7Pdzzvf7Pedr7o6IiCRXKu4AREQkXkoEIiIJp0QgIpJwSgQiIgmnRCAiknBKBCIiCadEIIlgZvPNzM2sKo9tLzSzX01GXCKlQIlASo6ZvWJmfWbWPGr5M2FhPj+eyEQqkxKBlKqXgfMHZ8zsGKAhvnBKQz5XNCL7SolAStX3gT+PzH8UuD26gZlNM7PbzazLzF41sy+YWSpclzaz681sk5mtAc4cY99/M7NOM1tvZl82s3Q+gZnZfWaWMbNtZvaomS2JrKs3s6+H8Wwzs1+ZWX247g/M7HEz22pm68zswnD5UjP7eOQYI6qmwqugT5jZ74Hfh8u+FR5ju5k9ZWZvj2yfNrPPmdlLZrYjXD/PzG4ys6+P+i4PmNkn8/neUrmUCKRU/RqYamaLwgL6POCOUdvcCEwD3gC8gyBxfCxcdzHwXuB4oA344Kh9bwP6gcPDbf4Y+Dj5+SmwEDgIeBq4M7LueuBE4G3ATOBvgZyZHRbudyPQArwJWJ7n5wG8H3gLsDicfzI8xkzg34H7zKwuXHc1wdXUe4CpwF8A3cD3gPMjybIZOC3cX5LM3fXSq6RewCsEBdQXgK8ApwMPA1WAA/OBNNAHLI7s95fA0nD6l8ClkXV/HO5bBcwGeoH6yPrzgUfC6QuBX+UZ6/TwuNMITqx2A8eNsd1ngf8c5xhLgY9H5kd8fnj8d00Qx5bBzwVeAM4eZ7vngXeH05cBD8b9761X/C/VN0op+z7wKLCAUdVCQDNQDbwaWfYqcEg4fTCwbtS6QYeF+3aa2eCy1KjtxxRenfwjcC7BmX0uEk8tUAe8NMau88ZZnq8RsZnZp4GLCL6nE5z5Dzau7+2zvgdcQJBYLwC+dQAxSYVQ1ZCULHd/laDR+D3Af4xavQnIEhTqgw4F1ofTnQQFYnTdoHUEVwTN7j49fE119yVM7E+BswmuWKYRXJ0AWBhTD/DGMfZbN85ygF2MbAifM8Y2Q48JDtsD/hb4EDDD3acD28IYJvqsO4Czzew4YBHww3G2kwRRIpBSdxFBtciu6EJ3HwDuBf7RzJrCOvirGW5HuBe4wsxazWwGcE1k307g58DXzWyqmaXM7I1m9o484mkiSCKbCQrv/x05bg64FfiGmR0cNtq+1cxqCdoRTjOzD5lZlZnNMrM3hbsuB84xswYzOzz8zhPF0A90AVVm9kWCK4JB3wH+wcwWWuBYM5sVxthB0L7wfeAH7r47j+8sFU6JQEqau7/k7svGWX05wdn0GuBXBI2et4brvg08BDxL0KA7+oriz4EaYBVB/fr9wNw8QrqdoJppfbjvr0et/zSwgqCwfR34JyDl7msJrmw+FS5fDhwX7vPPBO0dGwiqbu5k7x4Cfga8GMbSw8iqo28QJMKfA9uBfwPqI+u/BxxDkAxEMHcNTCOSJGb2hwRXToe5CgBBVwQiiWJm1cCVwHeUBGSQEoFIQpjZImArQRXYN2MOR0qIqoZERBJOVwQiIglXdjeUNTc3+/z58+MOQ0SkrDz11FOb3L1lrHVllwjmz5/PsmXj9SYUEZGxmNmr461T1ZCISMIpEYiIJJwSgYhIwpVdG8FYstksHR0d9PT0xB1K0dXV1dHa2kp1dXXcoYhIhaiIRNDR0UFTUxPz588n8ljhiuPubN68mY6ODhYsWBB3OCJSIYpWNWRmt5rZRjN7bpz1ZmY3mNlqM2s3sxP297N6enqYNWtWRScBADNj1qxZibjyEZHJU8w2gtsIRpYazxkEw/0tBC4B/vVAPqzSk8CgpHxPEZk8RasacvdHzWz+XjY5G7g9fPDVr81supnNDZ8VLyIVKBgaMRhlJzc0HbwzYlm4LeC5YJucB8tykX3cw32ix44sGzre0DbDy3Lhh0bnB7cZ3D6XG3mcXPj5uXCIx1xueFk0tuH54W3G3N9H7x9dH3x+dPtTF83muHnTC/7vEmcbwSGMfIZ6R7hsj0RgZpcQXDVw6KGHjl4du82bN3PqqacCkMlkSKfTtLQEN/D99re/paamZtx9ly1bxu23384NN9wwKbEeqIGcszs7QE92gN19A/T2D7C7L0dPfzA/uC545Ya3zQ7Q07fnst5sbugPEiB6wWOMmBkhOjvePiOW57FNtFCKFkiMLmxgj8JleL9gefR4uci2I481skAb2iZP+/KYsH07MkPfPVpwu+/5XcYq0MfbTw6MGRw0ta7iEkHe3P0W4BaAtra2kvtfatasWSxfvhyAa6+9lsbGRj796U8Pre/v76eqauyfuq2tjba2tqLFlh3IsaW7j63dWV7f1cfW7j5e35VlS3cfO3v7RxbafQNDBfpYBXlvNkffQG7iDx1DTVWK+uo09dVp6qpT1FWnw1eK6lRQQxkt2EZMjyrERq6LzA+N8z5yn9HbDy/3EcuNoOotZWGyCP4jlTKqzLBwuVmwXbD9yP0YXB+uS0X2Y3B+1H4jjsvI5DQRG50h97btPtYqDscW+S4MV0/a6O8TWR/dzwh/m8i2o/eD6G/FHr91KvJvEf2MsT4/FV036t8qZWPvn0rt+W8UjSeVMlKDcYTvwfzIz9rrNqnhZRZZt7djpkZ9j2KJMxGsZ+SYsq0Mjzdb9i688ELq6up45plnOOWUUzjvvPO48sor6enpob6+nu9+97sceeSRLF26lOuvv56f/OQnXHvttaxdu5Y1a9awdu1arrrqKq644oqhY+ZyzkDOyQ7keHz1Jl7v7mNLd5atu/p4fXRh393H1l1ZdvT2jxtjddqGCuTRhfT0hpqh+fqaNLVVaeprhrepr05TO7RPuH9Nami7uujxqtKkUmrbEClVcSaCB4DLzOxu4C3AtkK0D/z9j1ey6rXtBxxc1OKDp/J378tnXPOROjo6ePzxx0mn02zfvp3HHnuMqqoqHn74YT772c9x97330ZsdoD/nbO3uo7uvnxUrV3HPAz9j+7btvPPkN3HaOReQSlfRn/OhKpQN23u5+I7fjPisptoqpk+pZkZDDTMaaljQPIUZU4Lp4L2amQ01TG+oYeaUGqY3VFNXnS7I7yMi5a1oicDM7gL+CGg2sw7g74BqAHe/GXiQYAzX1UA38LFixRKXc889F8fY2ZNlzfqNXPPpq1nz0moco78/y+8y2+nYupvuvn7Wvt7Njp5+Tn7HaeweMOqnTmdmcwvbXt9Ea2sr6ZRRlTLSKaN/cw13X3JyWMhXM72+hpoq3SQuIvunmL2Gzp9gvQOfKPTn7s+ZeyG4O/05pzc7gPdk2dnbz9Y+Y1VncHXyd1/8Iiec/Afc/L272PjaOj74vv9F64x61k6tY0pNFUfMbqKlsZampkaWHDwNgPqaauZOrWHezIYRn7WxJs0Jb5g16d9RRCpTWTQWlxp3p68/2pAaNLT253Js3Z2lwfoZyDk1VSnmTKujvjoN2W5OXPxG5jdP4bZ/uZuUGTOn1NJYW0U6FdTVp1Km+wREZNIpEUwg507vGN0gB8L6esOorU7RVFdFfU2aGQ01TJ9Wx+b6alqa6jioqQ6Aaz7zGT760Y/y5S9/mTPPPDPOryQiMkLZjVnc1tbmowemef7551m0aNEBHzsX7SM/WOhnc0PdDFNmYW+Y1FDPmDh6xBTq+4pIcpjZU+4+Zl/1xF4R9A/khqt1wrP83v6BoX7m6ZRRX52meUoNdWG3ydqqlKpuRKTiJCYR7O7rZ3tP/9AdsdEbo6rTQb/4qfXVYV/5FNVpFfoikgyJSQQ7ewfYsL2H2qqgWmdmdc1Q9U51Wl0vRSS5EpMIZk6pZuaUatIpFfoiIlGJSQRKACIiY1PpKCKScIm5IiimA3kMNcDSpUupqanhbW97W9FjFZlQdjcM9EHt1H1/ZKmUJSWCApjoMdQTWbp0KY2NjUoEUnju0LcTdm2C7s3h+6bI++Y957O7gn2r6qFpNjTNhaY50DgneB96zYXG2VA3TQmjzCkRFMlTTz3F1Vdfzc6dO2lubua2225j7ty53HDDDdx8881UVVWxePFirrvuOm6++WbS6TR33HEHN954I29/+9vjDl9KVS4HPVv3XqgPrQvfB3rHPlZVHTQ0w5RZwfusw2FKMzTMgnQN7NwAOzLBe2YF7Hg4SCp7HEcJo9xVXiL46TXB/7SFNOcYOOO6vDd3dy6//HJ+9KMf0dLSwj333MPnP/95br31Vq677jpefvllamtr2bp1K9OnT+fSSy/d56sIKXP9vdC7E/p2hO+7wukd45y9Rwp3Hxj7mDWNQSE+pTkofOccMzzf0Bx5Dwv+min7XjD37oAdG2BnJkgSOzrD93wTRiRBNEYSxWAiKXTCcIeBLOSykOuHgf7gPZcNlw9EpsN1Q9PZ4e3T1VDdEL7qoaZh5HyqvB/pXnmJoAT09vby3HPP8e53vxuAgYEB5s6dC8CxS47iz849m/e//094/7l7fUBrMmz8HfRsg1QVpKsgVR2ZDufT1cEf2tB01eT/4Q0V3OFrj0J8Z1BIDq3bOXJ69LJcduLPrJs2XIDPWACtbWMX6IPz1XXF/x1qm4JX8+F7365QCaNmysgCe0RBPdZ0uO3gtO/fiHr7rKouSAjVU8ZOFDXh8jGXjbVPQzgfrk9XF/VqqvISwT6cuReLu7NkyRKeeOKJPdb91/dv4tHHHuPHDz/KP37ln1jxm/8O/sdNkt4dsOI+eOo26Hx2Pw9ikaRQPSpxRKbHTDCD04P7h4kl2x0psHeNLOzzKbghOFZNY1BY1jRCbWPwB984e3hZzZRweVP43ji8bW1TcBbfMCuIr1wVImHsyED36+G/U3VQ2A5Op9J7+fcffeIw+iSiauS+Q8ca7/+XquDKINs9/OqLTGd3B/+/ZHfvub5nW/CdhpbtDtpg9jVBWTpIDqd/BU74yP7/u4yj8hJBCaitraWrq4snnniCt771rWSzWV588UUWHXUU69a+zDvffQZ/cPo53L1wCTs3vEwTu9i+cVvwR1HTWLn1qOufDgr/FfcHfwyzj4b3XA8z3zDBJfo4l+tjXt5PcJz+Psh173nmmBsI/tDGKrgHl0UL7tEF+OC6dE3l/vsVQ74Jo5K4B72yxkseeySX8L2vG5oXFiUkJYIiSKVS3H///VxxxRVs27aN/v5+rrrqKo5YcCgXXPZ5tu3qxS3FFVd9kukL38L7zjY++JGL+NF/PcSNX/kCbz/1dKifWfb1jsCeZ/9V9XD0B6DtY3DIiSo0JXnMoKo2eJUIPYZ6Mu3eAltegeYjg/q/qNwA7N4K3V1B9rcUNMyEhpY96n7L4vu+9szw2X/fTjhoSVD4H/uhoO5bRCaVHkNdKrLdgI3dqJdKh41/M4PtdnUF3QF3bQqqH6a0lH73u94dQcH/1HdHnv2feGHQ0FnKsYskmBLBZMruDhq8bC9P9jAL66OnwNTscLfBLS8HDVhTmkuvcXmss//3XA/HnAv10+OOTkQmUDGJwN1Le/wA9yAR1E7Nf590ddjfenbQ+2DXJnz7a7C9C37wTXjzxTDvpHjOtIfO/m+DzuXh2f85cOLHdPYvUmYqIhHU1dWxefNmZs2aVbrJYLAPdHX9vu9rBvXT8bppbN7YSV2qE158KGiEnXMMnHQJHP3BPdsdiuG15UHVz9DZ/2I442tB3b/O/kXKUkU0FmezWTo6Oujp6Ykpqjxkdwf1/o2zD6i3QF1dHa2trVTneqH9HnjyO7BxFdRNh+MvgLa/gFlvLGDgBGf/z/0gOPt/7ZnI2f+F0Ppmnf2LlIG9NRZXRCIoC49+DX75ZbhmHdTtQ/XQRNzh1cfhyW/D8z8OrjoOPy2oNlr47gPrgvra8rDu/77hs/8TP6azf5EypF5DpSCzInhMQCGTAARn4/NPCV7bO4OC+6nb4K4Pw/TD4M0XwfEfCXoj5aN3Jzx3f+Tsvw6WnBN0/dTZv0hF0hXBZPnWm4L6/A9/v/ifNZANrg6e/A68+j9BYX70B+DNH4dDThh7n85ng8K//b7gsQoHLQ6qfo79ENTPKH7MIlJUuiKIW8+2oPvn8X82OZ+Xrg7q8I8+BzasDBLCs/fA8juDu3nffDEs+ZOgGum5HwSNvzr7F0ksJYLJsGFl8D7nuMn/7NlL4L3/DKddC8vvCpLCDy+Fhz4XXDn07YCWRXDGV3X2L5JQSgSTobM9eJ9zTHwx1E2Dky+Ft/wlrFkKT98edGU94aPx3YsgIiVBiWAyZNqDR0Q0zYk7kqDAf+M7g5eICLCXZx1IwWTaYc6xOusWkZKkRFBs/X3BKFxxVguJiOyFEkGxdf0ueLzE3GPjjkREZExKBMWWGWwojqHHkIhIHpQIiq2zPRh8euYb4o5ERGRMRU0EZna6mb1gZqvN7Jox1h9qZo+Y2TNm1m5m7ylmPLHIrIA5R0NKOVdESlPRSiczSwM3AWcAi4HzzWzxqM2+ANzr7scD5wH/p1jxxCKXCxOBGopFpHQV8zT1JGC1u69x9z7gbuDsUds4MPgUtmnAa0WMZ/JteTm4c3eOGopFpHQVMxEcAqyLzHeEy6KuBS4wsw7gQeDysQ5kZpeY2TIzW9bV1VWMWIsjsyJ4V48hESlhcVdcnw/c5u6twHuA75vtOaCvu9/i7m3u3tbS0jLpQe63TDtYOniWj4hIiSpmIlgPzIvMt4bLoi4C7gVw9yeAOqC5iDFNrswKaDkKquvijkREZFzFTARPAgvNbIGZ1RA0Bj8wapu1wKkAZraIIBGUUd3PBDrbVS0kIiWvaInA3fuBy4CHgOcJegetNLMvmdlZ4WafAi42s2eBu4ALvdxGyhnPzo2wM6MeQyJS8or69FF3f5CgETi67IuR6VXAKcWMITZDdxTrikBESlvcjcWVa2gMgqPjjUNEZAJKBMWSaYfph2rELxEpeUoExZJZoWohESkLSgTF0LsTNr+kRCAiZUGJoBg2PAe4uo6KSFlQIiiGwUdL6IpARMqAEkExdD4L9TNh6sFxRyIiMiElgmLIrAiqhTRYvYiUASWCQhvIwsZVqhYSkbKhRFBoXS/AQJ8SgYiUDSWCQtMYBCJSZpQICi3TDlX1MOvwuCMREcmLEkGhdbbD7CWQSscdiYhIXpQICsl9uMeQiEiZUCIopK2vQu82jUEgImVFiaCQhu4oPi7eOERE9oESQSF1toOlYPbiuCMREcmbEkEhZdqh+Qioro87EhGRvCkRFJLGIBCRMqREUCi7NsP29WooFpGyo0RQKJlng3d1HRWRMqNEUCgag0BEypQSQaF0tsPUVmiYGXckIiL7RImgUHRHsYiUKSWCQujrhs2/V7WQiJQlJYJC2LASPKceQyJSlpQICiHTHryrakhEypASQSFk2qFuOkybF3ckIiL7TImgEDrbg2ohDVYvImVowkRgZu8zMyWM8Qz0B4PVz9UTR0WkPOVTwH8Y+L2ZfdXMjip2QGVn8++hv0cNxSJStiZMBO5+AXA88BJwm5k9YWaXmFlT0aMrB51hQ7G6jopImcqrysfdtwP3A3cDc4E/AZ42s8uLGFt5yLRDujZ4/LSISBnKp43gLDP7T2ApUA2c5O5nAMcBnypueGUg0x4MRJOuijsSEZH9kk/p9QHgn9390ehCd+82s4uKE1aZGBysftFZcUciIrLf8qkauhb47eCMmdWb2XwAd//F3nY0s9PN7AUzW21m14yzzYfMbJWZrTSzf8878lKwrQN2b9GNZCJS1vJJBPcBucj8QLhsr8wsDdwEnAEsBs43s8WjtlkIfBY4xd2XAFflGXdpyKihWETKXz6JoMrd+wZnwumaPPY7CVjt7mvCfe4Gzh61zcXATe6+JTz2xvzCLhGZFYDB7CVxRyIist/ySQRdZjZUCW5mZwOb8tjvEGBdZL4jXBZ1BHCEmf2Pmf3azE4f60Bhd9VlZrasq6srj4+eJJ3tMOtwqJkSdyQiIvstn8biS4E7zexfACMo3P+8gJ+/EPgjoBV41MyOcfet0Y3c/RbgFoC2tjYv0GcfuEw7zDsp7ihERA7IhInA3V8CTjazxnB+Z57HXg9En8LWGi6L6gB+4+5Z4GUze5EgMTyZ52fEp/t12LYO3vzxuCMRETkgeXV+N7MzgSVAnYUPVnP3L02w25PAQjNbQJAAzgP+dNQ2PwTOB75rZs0EVUVr8o4+TkNjFOvREiJS3vK5oexmgucNXU5QNXQucNhE+7l7P3AZ8BDwPHCvu680sy9F2hweAjab2SrgEeBv3H3zfn2TyTaYCPSwOREpc/lcEbzN3Y81s3Z3/3sz+zrw03wO7u4PAg+OWvbFyLQDV4ev8pJph6aDYUpz3JGIiByQfHoN9YTv3WZ2MJAleN5Qsg2OQSAiUubySQQ/NrPpwNeAp4FXgPK6A7jQsrth04u6o1hEKsJeq4bCAWl+EXbn/IGZ/QSoc/dtkxJdqdq4CnxAdxSLSEXY6xWBu+cIHhMxON+b+CQAkTEIVDUkIuUvn6qhX5jZB8w0IO+QzAqonQoz5scdiYjIAcsnEfwlwUPmes1su5ntMLPtRY6rtGU0WL2IVI58hqpscveUu9e4+9RwfupkBFeScgOwYaXaB0SkYkx4H4GZ/eFYy0cPVJMYm1+CbLd6DIlIxcjnhrK/iUzXETxe+ingXUWJqNRl1FAsIpUln4fOvS86b2bzgG8WLaJSl2mHdA20HBV3JCIiBZFPY/FoHcCiQgdSNjrb4aBFkK6OOxIRkYLIp43gRmBwDIAU8CaCO4yTxz24IjjyjLgjEREpmHzaCJZFpvuBu9z9f4oUT2nb0Qndm2GOnjgqIpUjn0RwP9Dj7gMQDEpvZg3u3l3c0EqQ7igWkQqU153FQH1kvh74f8UJp8Rl2gGDOUfHHYmISMHkkwjqosNThtMNxQuphGXaYeYboLYp7khERAomn0Swy8xOGJwxsxOB3cULqYRpDAIRqUD5tBFcBdxnZq8RDFU5h2DoymTZvRW2vgonfjTuSERECiqfG8qeNLOjgCPDRS+4e7a4YZWgDc8F7+oxJCIVJp/B6z8BTHH359z9OaDRzP66+KGVGPUYEpEKlU8bwcXhCGUAuPsW4OLihVSiMiugcTY0zY47EhGRgsonEaSjg9KYWRqoKV5IJSrTrkdPi0hFyicR/Ay4x8xONbNTgbuAnxY3rBLT3wtdv1O1kIhUpHx6DX0GuAS4NJxvJ+g5lBwbn4dcv8YgEJGKlM8IZTngN8ArBGMRvAt4vrhhlZihMQiUCESk8ox7RWBmRwDnh69NwD0A7v7OyQmthGRWQE0TzFgQdyQiIgW3t6qh3wGPAe9199UAZvbJSYmq1HS2B88XSu3P8A0iIqVtbyXbOUAn8IiZfTtsKLa9bF+ZcrngZjI1FItIhRo3Ebj7D939POAo4BGCR00cZGb/amZ/PFkBxm7Ly9C3U+0DIlKx8mks3uXu/x6OXdwKPEPQkygZOp8N3tVjSEQq1D5Verv7Fne/xd1PLVZAJSfTDqkqDVYvIhVLrZ8TyayAlkVQVRt3JCIiRaFEMBGNQSAiFU6JYG92ZGDXRrUPiEhFUyLYm8yK4F09hkSkghU1EZjZ6Wb2gpmtNrNr9rLdB8zMzaytmPHss8EeQxqsXkQqWNESQfi46puAM4DFwPlmtniM7ZqAKwmeZ1RaMitgxnyomxZ3JCIiRVPMK4KTgNXuvsbd+4C7gbPH2O4fgH8CeooYy/7RGAQikgDFTASHAOsi8x3hsiFmdgIwz93/a28HMrNLzGyZmS3r6uoqfKRj6dkOr69RIhCRihdbY7GZpYBvAJ+aaNvwJrY2d29raWkpfnAAG1YG7+oxJCIVrpiJYD0wLzLfGi4b1AQcDSw1s1eAk4EHSqbBWGMQiEhCFDMRPAksNLMFZlYDnAc8MLjS3be5e7O7z3f3+cCvgbPcfVkRY8pfZzs0NENTsgZjE5HkKVoicPd+4DLgIYIRze5195Vm9iUzO6tYn1swmfagWsiS9+RtEUmWfMYs3m/u/iDw4KhlXxxn2z8qZiz7pL8vGKf4rX8ddyQiIkWnO4vHsukFyGXVPiAiiaBEMJbOsKF47nHxxiEiMgmUCMaSaYfqBpj5hrgjEREpOiWCsWRWwOyjIZWOOxIRkaJTIhgtlwsSgW4kE5GEUCIYbesr0Ltdg9GISGIoEYymMQhEJGGUCEbrbAdLw0F7PDFbRKQiKRGMllkBLUdCdV3ckYiITAolgtE0BoGIJIwSQdTOLtjRqYZiEUkUJYKowUdPq+uoiCSIEkHU0BgEuiIQkeRQIojqbIdph0L9jLgjERGZNEoEUbqjWEQSSIlgUO9O2LxaPYZEJHGUCAZtWAm42gdEJHGUCAapx5CIJJQSwaBMe9BIPPWQuCMREZlUSgSDMiuC9gENVi8iCaNEADCQhQ2rVC0kIomkRACw6UUY6FWPIRFJJCUC0BgEIpJoSgQQ3FFcVQ/NC+OORERk0ikRQNBjaPZiDVYvIomkROCuMQhEJNGUCLauhZ5tuqNYRBJLiWCwoXjucfHGISISEyWCTDtYSoPVi0hiKRF0tsOshVDTEHckIiKxUCLQGAQiknDJTgTdr8P2DvUYEpFES3Yi6Hw2eFePIRFJsGQnAvUYEhFJeiJoh6mt0DAz7khERGJT1ERgZqeb2QtmttrMrhlj/dVmtsrM2s3sF2Z2WDHj2UNmhaqFRCTxipYIzCwN3AScASwGzjez0Z31nwHa3P1Y4H7gq8WKZw993cHjp9VjSEQSrphXBCcBq919jbv3AXcDZ0c3cPdH3L07nP010FrEeEbauAo8pysCEUm8YiaCQ4B1kfmOcNl4LgJ+OtYKM7vEzJaZ2bKurq7CRDc4WL26jopIwoNXl+oAAAY+SURBVJVEY7GZXQC0AV8ba7273+Lube7e1tLSUpgP7WyHumkw/dDCHE9EpExVFfHY64F5kfnWcNkIZnYa8HngHe7eW8R4Rhp89LQGqxeRhCvmFcGTwEIzW2BmNcB5wAPRDczseOD/Ame5+8YixjLSQD9sWKlqIRERipgI3L0fuAx4CHgeuNfdV5rZl8zsrHCzrwGNwH1mttzMHhjncIW1eTX096jHkIgIxa0awt0fBB4cteyLkenTivn54xpqKFaPIRGRkmgsnnSZdkjXQvMRcUciIhK7ZCaCznY4aBGkq+OOREQkdslLBO4ag0BEJCJ5iWD7etj9unoMiYiEkpcIOnVHsYhIVPISQWYFYDB7SdyRiIiUhAQmgnaYdTjUNsYdiYhISUheIuhs1/0DIiIRyUoEu7fAtrXqMSQiEpGsRDA4RrEaikVEhigRiIgkXLISQWc7NM2FxgKNaSAiUgGSlQgyaigWERktOYkg2wNdL6haSERklOQkgo2rwAfUY0hEZJTkJAKNQSAiMqbkJIIpLXDkmTB9ftyRiIiUlKKOUFZSjjozeImIyAjJuSIQEZExKRGIiCScEoGISMIpEYiIJJwSgYhIwikRiIgknBKBiEjCKRGIiCScuXvcMewTM+sCXt3P3ZuBTQUMp9zp9xhJv8cw/RYjVcLvcZi7j/kM/rJLBAfCzJa5e1vccZQK/R4j6fcYpt9ipEr/PVQ1JCKScEoEIiIJl7REcEvcAZQY/R4j6fcYpt9ipIr+PRLVRiAiIntK2hWBiIiMokQgIpJwiUkEZna6mb1gZqvN7Jq444mLmc0zs0fMbJWZrTSzK+OOqRSYWdrMnjGzn8QdS9zMbLqZ3W9mvzOz583srXHHFBcz+2T4d/Kcmd1lZnVxx1QMiUgEZpYGbgLOABYD55vZ4nijik0/8Cl3XwycDHwiwb9F1JXA83EHUSK+BfzM3Y8CjiOhv4uZHQJcAbS5+9FAGjgv3qiKIxGJADgJWO3ua9y9D7gbODvmmGLh7p3u/nQ4vYPgj/yQeKOKl5m1AmcC34k7lriZ2TTgD4F/A3D3PnffGm9UsaoC6s2sCmgAXos5nqJISiI4BFgXme8g4YUfgJnNB44HfhNvJLH7JvC3QC7uQErAAqAL+G5YVfYdM5sSd1BxcPf1wPXAWqAT2ObuP483quJISiKQUcysEfgBcJW7b487nriY2XuBje7+VNyxlIgq4ATgX939eGAXkMg2NTObQVBzsAA4GJhiZhfEG1VxJCURrAfmReZbw2WJZGbVBEngTnf/j7jjidkpwFlm9gpBleG7zOyOeEOKVQfQ4e6DV4n3EySGJDoNeNndu9w9C/wH8LaYYyqKpCSCJ4GFZrbAzGoIGnweiDmmWJiZEdT/Pu/u34g7nri5+2fdvdXd5xP8f/FLd6/Is758uHsGWGdmR4aLTgVWxRhSnNYCJ5tZQ/h3cyoV2nBeFXcAk8Hd+83sMuAhgpb/W919ZcxhxeUU4CPACjNbHi77nLs/GGNMUlouB+4MT5rWAB+LOZ5YuPtvzOx+4GmC3nbPUKGPmtAjJkREEi4pVUMiIjIOJQIRkYRTIhARSTglAhGRhFMiEBFJOCUCkVHMbMDMlkdeBbuz1szmm9lzhTqeSCEk4j4CkX20293fFHcQIpNFVwQieTKzV8zsq2a2wsx+a2aHh8vnm9kvzazdzH5hZoeGy2eb2X+a2bPha/DxBGkz+3b4nPufm1l9bF9KBCUCkbHUj6oa+nBk3TZ3Pwb4F4KnlgLcCHzP3Y8F7gRuCJffAPy3ux9H8LyewbvZFwI3ufsSYCvwgSJ/H5G90p3FIqOY2U53bxxj+SvAu9x9Tfjgvoy7zzKzTcBcd8+GyzvdvdnMuoBWd++NHGM+8LC7LwznPwNUu/uXi//NRMamKwKRfePjTO+L3sj0AGqrk5gpEYjsmw9H3p8Ipx9neAjDPwMeC6d/AfwVDI2JPG2yghTZFzoTEdlTfeTJrBCM3zvYhXSGmbUTnNWfHy67nGBEr78hGN1r8GmdVwK3mNlFBGf+f0Uw0pVISVEbgUiewjaCNnffFHcsIoWkqiERkYTTFYGISMLpikBEJOGUCEREEk6JQEQk4ZQIREQSTolARCTh/j8D5XE7j4+naQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.92507004737854\n",
      "Training accuracy of best model:  0.99702597\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  56   60]\n",
      " [  21 1203]]\n",
      "the mean-f1 score: 0.77\n",
      "accuracy is: 0.94\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 47\n",
      "Percentage accuracy: 94.0\n",
      "total_count of train files: 187\n",
      "correctly predicted train files: 187\n",
      "Percentage accuracy: 100.0\n",
      "total_count of val files: 51\n",
      "correctly predicted val files: 47\n",
      "Percentage accuracy: 92.15686274509804\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "other"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_train='data_micro/data_m_other_TTV/train'\n",
    "data_dir_val= 'data_micro/data_m_other_TTV/val'\n",
    "data_dir_test='data_micro/data_m_other_TTV/test'\n",
    "pos_lab = 'other'\n",
    "neg_lab = 'No'\n",
    "labelfile = 'Labelother.csv'\n",
    "model_name = \"data_micro/DCBL_micro_other_jitter_mask.hdf5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FEYY86nqUvoy",
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#receive windowed training and validation data\n",
    "train_x, train_y, train_files = get_train_data(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "val_x, val_y, val_files = get_train_data(data_dir_val,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "test_x, test_y, test_files = get_train_data(data_dir_test,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x2, train_y2, all_counts = get_train_data_2(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)\n",
    "train_x3, train_y3, all_counts = get_train_data_3(data_dir_train,sub_dirs,pos_lab,neg_lab,labelfile)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "u2w_y9dmUvo1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5105, 12, 500)\n",
      "(5105,)\n",
      "(5099, 12, 500)\n",
      "(5099,)\n",
      "(5107, 12, 500)\n",
      "(5107,)\n",
      "(1373, 12, 500)\n",
      "(1373,)\n",
      "(1334, 12, 500)\n",
      "(1334,)\n"
     ]
    }
   ],
   "source": [
    "train_samples_2 = np.array(train_x2) \n",
    "train_labels2_2 = np.array(train_y2)\n",
    "print(train_samples_2.shape)\n",
    "print(train_labels2_2.shape)\n",
    "train_samples_3 = np.array(train_x3) \n",
    "train_labels2_3 = np.array(train_y3)\n",
    "print(train_samples_3.shape)\n",
    "print(train_labels2_3.shape)\n",
    "train_samples = np.array(train_x) \n",
    "train_labels2 = np.array(train_y)\n",
    "val_samples = np.array(val_x) \n",
    "val_labels2 = np.array(val_y)\n",
    "test_samples = np.array(test_x) \n",
    "test_labels2 = np.array(test_y)\n",
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)\n",
    "train_samples = np.vstack((train_samples,train_samples_2, train_samples_3))\n",
    "train_labels2 = np.hstack((train_labels2, train_labels2_2, train_labels2_3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15311, 12, 500)\n",
      "(15311,)\n",
      "(1373, 12, 500)\n",
      "(1373,)\n",
      "(1334, 12, 500)\n",
      "(1334,)\n"
     ]
    }
   ],
   "source": [
    "print(train_samples.shape)\n",
    "print(train_labels2.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels2.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15311, 12, 500)\n",
      "(15311, 2)\n",
      "(1373, 12, 500)\n",
      "(1373, 2)\n",
      "(1334, 12, 500)\n",
      "(1334,)\n"
     ]
    }
   ],
   "source": [
    "#convert to one hot encoding\n",
    "from keras.utils  import to_categorical\n",
    "train_labels = to_categorical(train_labels2)\n",
    "train_labels.shape\n",
    "val_labels = to_categorical(val_labels2)\n",
    "val_labels.shape\n",
    "test_labels = to_categorical(test_labels2)\n",
    "test_labels.shape\n",
    "print(train_samples.shape)\n",
    "print(train_labels.shape)\n",
    "print(val_samples.shape)\n",
    "print(val_labels.shape)\n",
    "print(test_samples.shape)\n",
    "print(test_labels2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [],
   "source": [
    "#For CNN and BidirLSTM:\n",
    "number_of_samples = 500\n",
    "train_samples = train_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "val_samples = val_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "test_samples = test_samples.reshape((-1, 12,number_of_samples, 1))\n",
    "#For LSTM, uncomment these lines and comment above two lines\n",
    "#train_samples = train_samples.reshape((-1, number_of_samples, 9))\n",
    "#val_samples = val_samples.reshape((-1, number_of_samples, 9))\n",
    "\n",
    "#9 refers to 9 accelerometer channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wf8LuDxlUvpH"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "500\n",
      "12\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_feat_map = 128\n",
    "num_hidden_lstm = 128\n",
    "num_classes = 2\n",
    "\n",
    "#Uncomment the classifier you are working with:\n",
    "#network_type = 'CNN'\n",
    "network_type = 'ConvLSTM'\n",
    "#network_type = 'LSTM'\n",
    "#for CNN/DCBL\n",
    "_, dim, win_len,_ = train_samples.shape\n",
    "#for LSTM uncomment the following line and comment above line\n",
    "#_, dim, win_len = train_samples.shape\n",
    "\n",
    "print(win_len)\n",
    "print(dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fAQRf1DNUvpK"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building the model ... \n",
      "ConvLSTM\n",
      "Model: \"sequential_11\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_61 (Conv2D)           (None, 12, 500, 128)      1408      \n",
      "_________________________________________________________________\n",
      "conv2d_62 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_41 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "conv2d_63 (Conv2D)           (None, 12, 500, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_42 (Batc (None, 12, 500, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_31 (MaxPooling (None, 12, 166, 128)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_64 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "conv2d_65 (Conv2D)           (None, 12, 166, 128)      163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_43 (Batc (None, 12, 166, 128)      512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_32 (MaxPooling (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_21 (Dropout)         (None, 12, 83, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_66 (Conv2D)           (None, 12, 83, 128)       163968    \n",
      "_________________________________________________________________\n",
      "batch_normalization_44 (Batc (None, 12, 83, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_33 (MaxPooling (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_22 (Dropout)         (None, 12, 41, 128)       0         \n",
      "_________________________________________________________________\n",
      "permute_11 (Permute)         (None, 41, 12, 128)       0         \n",
      "_________________________________________________________________\n",
      "reshape_11 (Reshape)         (None, 41, 1536)          0         \n",
      "_________________________________________________________________\n",
      "bidirectional_11 (Bidirectio (None, 256)               1704960   \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 2)                 514       \n",
      "=================================================================\n",
      "Total params: 2,528,770\n",
      "Trainable params: 2,527,746\n",
      "Non-trainable params: 1,024\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "p=0.5 #Dropout\n",
    "b = 1 #BatchNorm\n",
    "print('building the model ... ')\n",
    "model = Sequential()\n",
    "\n",
    "if network_type=='CNN' or network_type=='ConvLSTM':\n",
    "    model_conv(model, num_feat_map,p,b)\n",
    "    model_variant(model, num_feat_map, dim, network_type,p)\n",
    "if network_type=='LSTM':\n",
    "    model_LSTM(model,p)\n",
    "       \n",
    "model_output(model)    \n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "wrNCEAD8UvpL"
   },
   "outputs": [],
   "source": [
    "X_train = train_samples\n",
    "y_train = train_labels\n",
    "X_val = val_samples\n",
    "y_val = val_labels\n",
    "X_test = test_samples\n",
    "y_test = test_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tNsDOBNrUvpN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15311, 12, 500, 1) (1373, 12, 500, 1) (1334, 12, 500, 1)\n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape, X_val.shape, X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_true = np.argmax(y_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YpsJRPEfUvpP",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "epochs = 10\n",
    "\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# checkpoint\n",
    "filepath=model_name\n",
    "#checkpoint = ModelCheckpoint(filepath, monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n",
    "\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "#checkpoint2 = monitor_Training()\n",
    "\n",
    "callbacks_list = [checkpoint]\n",
    "\n",
    "H = model.fit(train_samples, train_labels,\n",
    "            batch_size=batch_size,\n",
    "            epochs=epochs,\n",
    "            verbose=1,\n",
    "            shuffle=True,\n",
    "            validation_data=(X_val, y_val),\n",
    "             callbacks=callbacks_list\n",
    "             )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "_ASUmtFSUvpQ"
   },
   "outputs": [],
   "source": [
    "history = H"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "C6yi5aaJUvpS"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum validation accuracy:  0.8069919943809509\n",
      "Training accuracy of best model:  0.99288094\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "print(np.argmax(np.array(history.history['val_accuracy'])))\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "print('Maximum validation accuracy: ',np.max(np.array(history.history['val_accuracy'])))\n",
    "print('Training accuracy of best model: ',np.array(history.history['accuracy'])[np.argmax(np.array(history.history['val_accuracy']))])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[212 136]\n",
      " [ 45 941]]\n",
      "the mean-f1 score: 0.81\n",
      "accuracy is: 0.86\n"
     ]
    }
   ],
   "source": [
    "model = load_model(model_name)\n",
    "y_pred = np.argmax(model.predict(X_test), axis=1)\n",
    "y_true = np.argmax(y_test, axis=1)\n",
    "cf_matrix = confusion_matrix(y_true, y_pred)\n",
    "print(cf_matrix)\n",
    "class_wise_f1 = np.round(f1_score(y_true, y_pred, average=None)*100)*0.01\n",
    "accuracy = accuracy_score(y_true, y_pred)\n",
    "\n",
    "\n",
    "print('the mean-f1 score: {:.2f}'.format(np.mean(class_wise_f1)))\n",
    "print('accuracy is: {:.2f}'.format(accuracy))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "total_count of test files: 50\n",
      "correctly predicted test files: 44\n",
      "Percentage accuracy: 88.0\n",
      "total_count of train files: 188\n",
      "correctly predicted train files: 187\n",
      "Percentage accuracy: 99.46808510638297\n",
      "total_count of val files: 50\n",
      "correctly predicted val files: 41\n",
      "Percentage accuracy: 82.0\n"
     ]
    }
   ],
   "source": [
    "# Test data accuracy:\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_test, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "# Train accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_train, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of train files:', total_count)\n",
    "print('correctly predicted train files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)\n",
    "#val accuracy\n",
    "total_count, correct_count = get_prediction_accuracy(data_dir_val, sub_dirs,pos_lab,neg_lab,labelfile,model)\n",
    "print('total_count of val files:', total_count)\n",
    "print('correctly predicted val files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [],
   "source": [
    "del train_x, train_y, train_files, val_x, val_y, val_files, test_x, test_y,test_files\n",
    "del train_x2, train_y2, all_counts\n",
    "del train_x3, train_y3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir_test='test_micro_ov/'\n",
    "labelfile = 'data/LabelTable.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/sandeep/anaconda3/envs/sandeep-env/lib/python3.7/site-packages/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "If using Keras pass *_constraint arguments to layers.\n",
      "WARNING:tensorflow:From /home/sandeep/anaconda3/envs/sandeep-env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4070: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
      "\n",
      "WARNING:tensorflow:From /home/sandeep/anaconda3/envs/sandeep-env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:422: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "modelcut = load_model('data_micro/DCBL_micro_Cut_jitter_mask.hdf5')\n",
    "modelwash = load_model('data_micro/DCBL_micro_Wash_jitter_mask.hdf5')\n",
    "modeltake = load_model('data_micro/DCBL_micro_Take_jitter_mask.hdf5')\n",
    "modelput = load_model('data_micro/DCBL_micro_Put_jitter_mask.hdf5')\n",
    "modelpeel = load_model('data_micro/DCBL_micro_Peel_jitter_mask.hdf5')\n",
    "modeladd = load_model('data_micro/DCBL_micro_ADD_jitter_mask.hdf5')\n",
    "modelmix = load_model('data_micro/DCBL_micro_Mix_jitter_mask.hdf5')\n",
    "modelpour = load_model('data_micro/DCBL_micro_Pour_jitter_mask.hdf5')\n",
    "modelopen = load_model('data_micro/DCBL_micro_Open_jitter_mask.hdf5')\n",
    "modelother = load_model('data_micro/DCBL_micro_other_jitter_mask.hdf5')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_test_prediction_accuracy(data_dir, sub_dirs,labelfile,modelcut,modelwash,modeltake,modelput,modelpeel,modeladd,modelmix,modelpour,modelopen,modelother):\n",
    "\n",
    "    files = os.listdir(data_dir+'/left_hip')\n",
    "    number_of_samples = 500\n",
    "\n",
    "    labels_micro =dict()\n",
    "    \n",
    "    labels_micro['Cut'] = 0\n",
    "    labels_micro['Wash'] = 1\n",
    "    labels_micro['Take'] = 2\n",
    "    labels_micro['Put'] = 3\n",
    "    labels_micro['Peel'] = 4\n",
    "    labels_micro['Add'] = 5\n",
    "    labels_micro['Mix'] = 6\n",
    "    labels_micro['Pour'] = 7\n",
    "    labels_micro['Open'] = 8\n",
    "    labels_micro['other'] = 9 \n",
    "    \n",
    "    \n",
    "    #read the labels\n",
    "    labels_loc = labelfile\n",
    "    file_label = open(labels_loc, newline='')\n",
    "    label_reader = csv.reader(file_label)\n",
    "    file_label_mapping = dict()\n",
    "    \n",
    "    for row in label_reader:\n",
    "        label = row[2].split(',')\n",
    "        \n",
    "        label_file_micro = [0]*10\n",
    "        for lab in label:\n",
    "            if lab!='' and lab !='micro':\n",
    "                label_file_micro[labels_micro[lab]] = 1\n",
    "                \n",
    "        \n",
    "        file_label_mapping[row[0]+'.csv'] = label_file_micro\n",
    "\n",
    "\n",
    "    total_count = 0\n",
    "    correct_count = 0\n",
    "\n",
    "    for f in files:\n",
    "\n",
    "        file_data = []\n",
    "        file_label = []\n",
    "\n",
    "        st_index = 0\n",
    "        end_index = 30000\n",
    "        step = 1000 #overlapping window, step: 1000. \n",
    "        window_index = 10000 #6 second window\n",
    "\n",
    "        print('reading file:',f)\n",
    "        f_name = f\n",
    "\n",
    "        if f_name == '.DS_Store':\n",
    "            continue\n",
    "\n",
    "        total_count = total_count+1\n",
    "\n",
    "        curr_label_file = file_label_mapping[f_name]\n",
    "\n",
    "        while st_index+step < end_index:\n",
    "\n",
    "            data, data_mask, data_count = parse_IMU_files(data_dir, sub_dirs, st_index, st_index+window_index,  f, number_of_samples)\n",
    "            st_index = st_index+step\n",
    "\n",
    "            if data_count<data_min_count:\n",
    "                continue\n",
    "\n",
    "            train_data_sample  = np.zeros((12, number_of_samples))\n",
    "            train_data_label   = curr_label_file\n",
    "            for i in range(len(data)):\n",
    "                for j in range(len(data[i])):\n",
    "                    train_data_sample[i*4,j]=data[i][j][0]\n",
    "                    train_data_sample[i*4+1,j]=data[i][j][1]\n",
    "                    train_data_sample[i*4+2,j]=data[i][j][2]\n",
    "                    \n",
    "                    #mask information\n",
    "                    train_data_sample[i*4+3,j]=data_mask[i][j]\n",
    "                    \n",
    "            file_data.append(train_data_sample)\n",
    "            #file_label.append(train_data_label)\n",
    "\n",
    "        file_data = np.array(file_data)\n",
    "        file_label = curr_label_file\n",
    "\n",
    "        file_data = file_data.reshape((-1,12,500, 1))\n",
    "\n",
    "        #y_pred = np.argmax(model.predict(file_data), axis=1) \n",
    "        \n",
    "        y_pred_cut = modelcut.predict(file_data)\n",
    "     \n",
    "        y_pred_wash = modelwash.predict(file_data)\n",
    "\n",
    "        y_pred_take = modeltake.predict(file_data)\n",
    "\n",
    "        y_pred_put = modelput.predict(file_data)\n",
    "\n",
    "        y_pred_peel = modelpeel.predict(file_data)\n",
    "  \n",
    "        y_pred_add = modeladd.predict(file_data)\n",
    " \n",
    "        y_pred_mix = modelmix.predict(file_data)\n",
    "    \n",
    "        y_pred_pour = modelpour.predict(file_data)\n",
    "  \n",
    "        y_pred_open = modelopen.predict(file_data)\n",
    "    \n",
    "        y_pred_other = modelother.predict(file_data)\n",
    "\n",
    "\n",
    "        \n",
    "        y_pred = np.zeros((file_data.shape[0],10))\n",
    "        \n",
    "        for j in range(0,y_pred.shape[0]):\n",
    "            if(y_pred_cut[j,0]>=y_pred_cut[j,1]):\n",
    "                y_pred[j,0] = 1\n",
    "            if(y_pred_wash[j,0]>=y_pred_wash[j,1]):\n",
    "                y_pred[j,1] = 1\n",
    "            if(y_pred_take[j,0]>=y_pred_take[j,1]):\n",
    "                y_pred[j,2] = 1\n",
    "            if(y_pred_put[j,0]>=y_pred_put[j,1]):\n",
    "                y_pred[j,3] = 1\n",
    "            if(y_pred_peel[j,0]>=y_pred_peel[j,1]):\n",
    "                y_pred[j,4] = 1\n",
    "            if(y_pred_add[j,0]>=y_pred_add[j,1]):\n",
    "                y_pred[j,5] = 1\n",
    "            if(y_pred_mix[j,0]>=y_pred_mix[j,1]):\n",
    "                y_pred[j,6] = 1\n",
    "            if(y_pred_pour[j,0]>=y_pred_pour[j,1]):\n",
    "                y_pred[j,7] = 1\n",
    "            if(y_pred_open[j,0]>=y_pred_open[j,1]):\n",
    "                y_pred[j,8] = 1\n",
    "            if(y_pred_other[j,0]>=y_pred_other[j,1]):\n",
    "                y_pred[j,9] = 1\n",
    "        \n",
    "        #print(y_pred.shape)\n",
    "        \n",
    "        #print(y_pred)\n",
    "        #y_pred = np.where(y_pred > 0, 1, 0)\n",
    "        \n",
    "        loc_arr = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0])\n",
    "        \n",
    "        \n",
    "        for i in range(y_pred.shape[0]):\n",
    "            loc_arr = loc_arr + y_pred[i]\n",
    "        \n",
    "        \n",
    "        loc_arr = loc_arr/y_pred.shape[0]\n",
    "        \n",
    "        max_val = loc_arr.max()\n",
    "        \n",
    "        loc_arr = np.where(loc_arr > max_val/3.0, 1, 0)\n",
    "        \n",
    "        #print(loc_arr)\n",
    "        \n",
    "        file_label = np.array(file_label)\n",
    "        \n",
    "        #print(file_label)\n",
    "\n",
    "        #correct prediction\n",
    "        \n",
    "        if np.array_equal(file_label, loc_arr):\n",
    "            correct_count = correct_count+1\n",
    "        \n",
    "        #print('*'*10)\n",
    "        #break\n",
    "    \n",
    "    return total_count, correct_count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Test data accuracy:\n",
    "total_count, correct_count = get_test_prediction_accuracy(data_dir_test, sub_dirs,labelfile,modelcut,modelwash,modeltake,modelput,modelpeel,modeladd,modelmix,modelpour,modelopen,modelother)\n",
    "print('total_count of test files:', total_count)\n",
    "print('correctly predicted test files:', correct_count)\n",
    "print('Percentage accuracy:', (correct_count/total_count)*100.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "Macro-Activity",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}