segnet_WH.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "b28a531e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting mock\n",
      "  Downloading mock-4.0.3-py3-none-any.whl (28 kB)\n",
      "Installing collected packages: mock\n",
      "Successfully installed mock-4.0.3\n"
     ]
    }
   ],
   "source": [
    "!pip install mock\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "id": "8e7b9ef5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "        <script type=\"text/javascript\">\n",
       "        window.PlotlyConfig = {MathJaxConfig: 'local'};\n",
       "        if (window.MathJax) {MathJax.Hub.Config({SVG: {font: \"STIX-Web\"}});}\n",
       "        if (typeof require !== 'undefined') {\n",
       "        require.undef(\"plotly\");\n",
       "        requirejs.config({\n",
       "            paths: {\n",
       "                'plotly': ['https://cdn.plot.ly/plotly-latest.min']\n",
       "            }\n",
       "        });\n",
       "        require(['plotly'], function(Plotly) {\n",
       "            window._Plotly = Plotly;\n",
       "        });\n",
       "        }\n",
       "        </script>\n",
       "        "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import enum\n",
    "import itertools\n",
    "import functools\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import scale\n",
    "from keras import backend as K\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import mock\n",
    "import torch.optim as optim\n",
    "#from tqdm import tqdm_notebook as tqdm\n",
    "from tqdm.notebook import trange, tqdm\n",
    "\n",
    "\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "\n",
    "import seaborn as sns\n",
    "# from visuals import plot_examples, plot_3d_example\n",
    "import matplotlib.gridspec as gridspec\n",
    "import plotly.graph_objs as go\n",
    "import plotly.offline as py\n",
    "%matplotlib inline\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.layers import Conv2D, Dropout, Flatten, MaxPooling2D, Dense\n",
    "from keras.layers import Activation, Dropout, Flatten, Dense\n",
    "import torchnet as tnt\n",
    "from functools import partial\n",
    "import torch.nn.functional as nnf\n",
    "\n",
    "\n",
    "py.init_notebook_mode(connected=True)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "id": "b66cf646",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bands: ['VV', 'VH', 'VV/VH']\n",
      "Reading pre-stored stack of patches... (32, 64, 64, 3)\n",
      "Reading pre-stored stack of GT patches... (32, 64, 64)\n"
     ]
    }
   ],
   "source": [
    "# here is where you save your images\n",
    "datafolder=r'C:\\Users\\vaak\\Desktop\\UoS\\weedwatch project\\S1_Vembanad\\STK'\n",
    "\n",
    "# list of convenience\n",
    "bands = ['VV', 'VH', 'VV/VH']\n",
    "print('bands:',bands)\n",
    "#%% Loading the images\n",
    "#[C11, C22_full, C12] = np.load(path_save + 'filename.npy')\n",
    "img=np.load(datafolder + '\\TrainingData_Vembanadlake_64.npy')\n",
    "print('Reading pre-stored stack of patches...',images_orig.shape)    \n",
    "\n",
    "GT=np.load(datafolder + '\\GTData_Vembanadlake_64.npy')\n",
    "print('Reading pre-stored stack of GT patches...',GT.shape)    \n",
    "\n",
    "\n",
    "category_names = [\n",
    " 'clean water',\n",
    " 'infested water']\n",
    "\n",
    "\n",
    "# load all data\n",
    "N_images = 16  # per category\n",
    "nb_classes = 2\n",
    "patch_size = 64  # width and height in pixel\n",
    "channels = 3\n",
    "nb_images = N_images * nb_classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "id": "a0308915",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25 tiles for training, 7 tiles for testing\n"
     ]
    }
   ],
   "source": [
    "shuffled_indices = np.arange(nb_images)\n",
    "np.random.seed(0)\n",
    "np.random.shuffle(shuffled_indices)\n",
    "\n",
    "training_indices = shuffled_indices[:int(0.8*nb_images)]               # 70% for training\n",
    "#val_indices = shuffled_indices[int(0.7*nb_images):int(0.8*nb_images)]  # 30% for val\n",
    "test_indices = shuffled_indices[int(0.8*nb_images):]                   # 20% for test\n",
    "\n",
    "train_obs = images_orig[training_indices]\n",
    "#val_images = images_orig[val_indices]\n",
    "test_obs = images_orig[test_indices]\n",
    "\n",
    "train_gt=GT[training_indices]\n",
    "test_gt=GT[test_indices]\n",
    "\n",
    "n_train = train_obs.shape[0]\n",
    "n_test = test_obs.shape[0]\n",
    "class_names = [\"clean water\", \"infested water\"]\n",
    "\n",
    "\n",
    "\n",
    "print(\"%d tiles for training, %d tiles for testing\" % (n_train, n_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "2a6248d1",
   "metadata": {},
   "outputs": [],
   "source": [
    "#[4] data loader\n",
    "\n",
    "def augment(obs, gt):\n",
    "  \"\"\"augmentation function\n",
    "  Leave untouched for now until question XX\n",
    "  \"\"\"\n",
    "  return obs, gt#does nothing, leaves like that for now\n",
    "  #random gaussian noise\n",
    "  sigma, clip= 0.01, 0.02 # https://github.com/charlesq34/pointnet/blob/master/provider.py#L74\n",
    "  obs = obs + np.clip(sigma * np.random.randn(*obs.shape), -1*clip, clip).astype(np.float32)\n",
    "  #random rotation 0 90 180 270 degree\n",
    "  #n_turn =  np.random.randint(4)   #removed because detrmental, see Q14\n",
    "  #obs = np.rot90(obs, k=n_turn, axes=(1,2)).copy()\n",
    "  #gt = np.rot90(gt, k=n_turn, axes=(0,1)).copy()\n",
    "  return obs, gt\n",
    "\n",
    "\n",
    "def tile_loader(tile_index, train = True):\n",
    "  \"\"\"\n",
    "  load a tile and returns the obseravtion and associated ground truth\n",
    "  INPUT:\n",
    "  tile_index = int, index of the tile\n",
    "  train = int, train = 1 iff in the train set\n",
    "  OUTPUT\n",
    "  obs, [256 x 256 x 4] float array containing the observation\n",
    "  gt, [256 x 256] uint8 array, containing the pixels semantic labels  \n",
    "  \"\"\"\n",
    "  if train:\n",
    "    obs = train_obs[tile_index,:,:,:].transpose(2,0,1) #put channels first\n",
    "    gt = train_gt[tile_index,:,:]\n",
    "  else:\n",
    "    obs = test_obs[tile_index,:,:,:].transpose(2,0,1)\n",
    "    gt = test_gt[tile_index,:,:]\n",
    "    \n",
    "  if train: #augmentation - for training only\n",
    "    obs, gt = augment(obs, gt)   \n",
    "  \n",
    "  obs = torch.from_numpy(obs)\n",
    "  gt = torch.from_numpy(gt)\n",
    "   \n",
    "  return obs, gt.long()\n",
    "\n",
    "#putting the dataset into the ListDataset wrapper\n",
    "test_set  = tnt.dataset.ListDataset(list(range(n_test)),functools.partial(tile_loader, train=False))\n",
    "train_set = tnt.dataset.ListDataset(list(range(n_train)),functools.partial(tile_loader, train=True))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "ed496537",
   "metadata": {},
   "outputs": [],
   "source": [
    "def viewer(n_shown = 3, category = 'cig', train = True, model = None, use_mask = False):\n",
    "  \"\"\" plot 'n_shown' random tiles train/test set with the following visuals:\n",
    "  if 'c' in category : rgb color\n",
    "  if 'i' in category : infrared\n",
    "  if 'g' in category : ground truth\n",
    "  if 'p' in category : prediction\n",
    "  if 'e' in category : error\n",
    "  Note that for 'p' or 'e' ou need to add a trained model as input\n",
    "  \n",
    "  \"\"\"\n",
    "  n_category = len(category) #number of types of image to show\n",
    "  fig = plt.figure(figsize=(n_category * 5, n_shown * 5)) #adapted dimension\n",
    "  \n",
    "  subplot_index = 1 #keep track of current subplot\n",
    "  \n",
    "  #chose random tiles\n",
    "  tile_indices = np.random.choice(n_train, n_shown) if train \\\n",
    "  else np.random.choice(n_test, n_shown)\n",
    "  \n",
    "  for tile_index in tile_indices:\n",
    "    \n",
    "    tile, gt = tile_loader(tile_index, train = train)\n",
    "    \n",
    "    if 'c' in category:\n",
    "      ax = fig.add_subplot(n_shown, n_category, subplot_index, aspect='equal')\n",
    "      if subplot_index <= n_category : \n",
    "        ax.set(title='RGB')\n",
    "      view_rgb(tile, ax = ax)\n",
    "      subplot_index += 1\n",
    "    if 'i' in category:  \n",
    "      ax = fig.add_subplot(n_shown, n_category, subplot_index, aspect='equal')\n",
    "      if subplot_index <= n_category : \n",
    "        ax.set(title='Infrared')\n",
    "      view_rgb(tile, ax = ax)\n",
    "      subplot_index += 1\n",
    "    if 'g' in category:  \n",
    "      ax = fig.add_subplot(n_shown, n_category, subplot_index, aspect='equal')\n",
    "      if subplot_index <= n_category : \n",
    "        ax.set(title='Ground Truth')\n",
    "      view_labels(gt, ax = ax)\n",
    "      subplot_index += 1\n",
    "    if 'p' in category:  \n",
    "      ax = fig.add_subplot(n_shown, n_category, subplot_index, aspect='equal')\n",
    "      if subplot_index <= n_category : \n",
    "        ax.set(title='Prediction')\n",
    "      pred = model(tile[None,:,:,:].cuda()).cpu().argmax(1).squeeze()\n",
    "      if use_mask:\n",
    "        pred[gt==99] = 99\n",
    "      view_labels(pred, ax = ax)\n",
    "      subplot_index += 1  \n",
    "    if 'e' in category: \n",
    "      ax = fig.add_subplot(n_shown, n_category, subplot_index, aspect='equal')\n",
    "      if subplot_index <= n_category : \n",
    "        ax.set(title='Error')\n",
    "      #pred = model(tile[None,:,:,:].cuda()).cpu().argmax(1).squeeze()\n",
    "      pred = model(tile[None,:,:,:].cpu()).argmax(1).squeeze()\n",
    "\n",
    "      view_error(pred, gt, ax = ax)\n",
    "      subplot_index += 1\n",
    "  plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "id": "ba5cdeeb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3, 64, 64, 3)"
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_obs[:3,:,].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "id": "1b5b3854",
   "metadata": {},
   "outputs": [],
   "source": [
    "def view_rgb(tile, ax = None):\n",
    "  \"\"\" show the rgb values of the tile in figure ax\"\"\"\n",
    "  if ax==None:\n",
    "    fig = plt.figure(figsize=(10, 10))\n",
    "    ax = fig.add_subplot(1, 1, 1, aspect='equal')\n",
    "  tile_corrected = np.minimum(np.maximum(tile[:3,:,], 0), 1) #normalization\n",
    "  ax.imshow(tile_corrected.transpose(0,2).\n",
    "  transpose(0,1)) #put channels back as dim 3\n",
    "  plt.axis('off')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "id": "971c2558",
   "metadata": {},
   "outputs": [],
   "source": [
    "def view_labels(label, ax = None, mask = None):\n",
    "  \"\"\" show the ground truth with a colorcode corresponding to labels\"\"\"\n",
    "  if ax==None:\n",
    "    fig = plt.figure(figsize=(10, 10))\n",
    "    ax = fig.add_subplot(1, 1, 1, aspect='equal')\n",
    "  if mask is not None:\n",
    "    label[mask] = 99\n",
    "  n_pixel = label.shape[1]\n",
    "  colors = np.zeros((n_pixel,n_pixel,3))\n",
    "  colors[np.where(label==99)] = [1  ,1  ,1  ] #not labelled\n",
    "  colors[np.where(label==0)]  = [1  ,0.8,0.8] #building limit\n",
    "  colors[np.where(label==1)]  = [0  ,0  ,1  ] #water\n",
    "  colors[np.where(label==2)]  = [0.9,0.9,0  ] #fields\n",
    "  colors[np.where(label==3)]  = [0.5,0.5,0.5] #road\n",
    "  colors[np.where(label==4)]  = [0  ,.8  ,0  ] #vegetation\n",
    "  colors[np.where(label==5)]  = [1,  0  ,0  ] #building\n",
    "  ax.imshow(colors)\n",
    "  plt.axis('off')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "id": "d18e1627",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x1800 with 15 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "viewer(n_shown = 5, category = 'cig', train = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "id": "c2917c8d",
   "metadata": {},
   "outputs": [],
   "source": [
    "#[7]\n",
    "class ConfusionMatrix:\n",
    "  def __init__(self, n_class, class_names):\n",
    "    self.CM = np.zeros((n_class, n_class))\n",
    "    self.n_class = n_class\n",
    "    self.class_names = class_names\n",
    "  \n",
    "  def clear(self):\n",
    "    self.CM = np.zeros((self.n_class, self.n_class))\n",
    "    \n",
    "  def add_batch(self, gt, pred):\n",
    "    self.CM +=  confusion_matrix(gt, pred, labels = list(range(self.n_class)))\n",
    "    \n",
    "  def overall_accuracy(self):#percentage of correct classification\n",
    "    return 100*self.CM.trace() / self.CM.sum()\n",
    "\n",
    "  def class_IoU(self, show = 1):\n",
    "    ious = np.diag(self.CM)/ (self.CM.sum(0) + self.CM.sum(1) - np.diag(self.CM))\n",
    "    if show:\n",
    "      print('  |  '.join('{} : {:3.2f}%'.format(name, 100*iou) for name, iou in zip(self.class_names,ious)))\n",
    "    #do not count classes that are not present in the dataset in the mean IoU\n",
    "    return 100*np.nansum(ious) / (np.logical_not(np.isnan(ious))).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "id": "57c1bd6a",
   "metadata": {},
   "outputs": [],
   "source": [
    "#[9]\n",
    "class SegNet(nn.Module):\n",
    "  \"\"\"\n",
    "  SegNet network for semantic segmentation\n",
    "  \"\"\"\n",
    "  \n",
    "  def __init__(self, n_channels, conv_width,  n_class, cuda = 1):\n",
    "    \"\"\"\n",
    "    initialization function\n",
    "    n_channels, int, number of input channel\n",
    "    conv_width, int list, depth of the convs\n",
    "    n_class = int,  the number of classes\n",
    "    \"\"\"\n",
    "    super(SegNet, self).__init__() #necessary for all classes extending the module class\n",
    "    self.is_cuda = cuda\n",
    "    \n",
    "    assert((conv_width[3] == conv_width[5]) \\\n",
    "     and (conv_width[1] == conv_width[7]))\n",
    "    \n",
    "    self.maxpool=nn.MaxPool2d(2,2,return_indices=True) #maxpooling layer\n",
    "    self.unpool=nn.MaxUnpool2d(2,2) #unpooling layer\n",
    "    #encoder\n",
    "    self.c0 = nn.Sequential(nn.Conv2d(n_channels,conv_width[0],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[0]),nn.ReLU(True))\n",
    "    self.c1 = nn.Sequential(nn.Conv2d(conv_width[0],conv_width[1],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[1]),nn.ReLU(True))\n",
    "    self.c2 = nn.Sequential(nn.Conv2d(conv_width[1],conv_width[2],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[2]),nn.ReLU(True))\n",
    "    self.c3 = nn.Sequential(nn.Conv2d(conv_width[2],conv_width[3],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[3]),nn.ReLU(True))\n",
    "    self.c4 = nn.Sequential(nn.Conv2d(conv_width[3],conv_width[4],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[4]),nn.ReLU(True))\n",
    "    self.c5 = nn.Sequential(nn.Conv2d(conv_width[4],conv_width[5],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[5]),nn.ReLU(True))\n",
    "    #decoder\n",
    "    self.c6=nn.Sequential(nn.Conv2d(conv_width[5]+conv_width[3],conv_width[6],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[6]),nn.ReLU(True))\n",
    "    self.c7=nn.Sequential(nn.Conv2d(conv_width[6],conv_width[7],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[7]),nn.ReLU(True))       \n",
    "    self.c8=nn.Sequential(nn.Conv2d(conv_width[1] + conv_width[7],conv_width[8],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[8]),nn.ReLU(True))\n",
    "    self.c9=nn.Sequential(nn.Conv2d(conv_width[8],conv_width[9],3,padding=1, padding_mode='reflect'),nn.BatchNorm2d(conv_width[9]),nn.ReLU(True))\n",
    "    #final classifying layer\n",
    "    self.c10=nn.Conv2d(conv_width[9],n_class,1,padding=0, padding_mode='reflect')\n",
    "\n",
    "    #weight initialization\n",
    "\n",
    "    self.c0[0].apply(self.init_weights)\n",
    "    self.c0[0].apply(self.init_weights)\n",
    "    self.c2[0].apply(self.init_weights)\n",
    "    self.c3[0].apply(self.init_weights)\n",
    "    self.c4[0].apply(self.init_weights)\n",
    "    self.c5[0].apply(self.init_weights)\n",
    "    self.c6[0].apply(self.init_weights)\n",
    "    self.c7[0].apply(self.init_weights)\n",
    "    self.c8[0].apply(self.init_weights)\n",
    "    self.c9[0].apply(self.init_weights)\n",
    "    self.c10.apply(self.init_weights)\n",
    "    \n",
    "    if cuda: #put the model on the GPU memory\n",
    "      self.cuda()\n",
    "    \n",
    "  def init_weights(self,layer): #gaussian init for the conv layers\n",
    "    nn.init.kaiming_normal_(layer.weight, mode='fan_out', nonlinearity='relu')\n",
    "    \n",
    "  def forward(self,input):\n",
    "    \"\"\"\n",
    "    the function called to run inference\n",
    "    \"\"\"  \n",
    "    if self.is_cuda: #put data on GPU\n",
    "        input = input.cuda()\n",
    "    #encoder\n",
    "    #level a\n",
    "    x1 = self.c1(self.c0(input))\n",
    "    x2, indices_a_b =self.maxpool(x1)\n",
    "    #level b\n",
    "    x3=self.c3(self.c2(x2))\n",
    "    x4, indices_b_c =self.maxpool(x3)\n",
    "    #level c\n",
    "    x5 = self.c5(self.c4(x4))\n",
    "    #decoder\n",
    "    #level b       \n",
    "    x6 = self.unpool(x5, indices_b_c)\n",
    "    x7 = self.c7(self.c6(torch.cat((x6,x3),1)))\n",
    "    #level a       \n",
    "    x8 = self.unpool(x7, indices_a_b)\n",
    "    x9 = self.c9(self.c8(torch.cat((x8,x1),1)))\n",
    "    #output         \n",
    "    out = self.c10(x9)\n",
    "    return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "id": "4093fbd8",
   "metadata": {},
   "outputs": [
    {
     "ename": "AssertionError",
     "evalue": "Torch not compiled with CUDA enabled",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-134-942802ea3409>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;31m#we consider the first point cloud from the training set\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0mtile\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgt\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtile_loader\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 5\u001b[1;33m \u001b[0msegnet\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mSegNet\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m3\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m16\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m16\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m32\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m32\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m64\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m32\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m32\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m16\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m32\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m16\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m6\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      6\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msegnet\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      7\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'Total number of parameters: {}'\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msum\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnumel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mp\u001b[0m \u001b[1;32min\u001b[0m \u001b[0msegnet\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-133-e241b1e8a9a2>\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, n_channels, conv_width, n_class, cuda)\u001b[0m\n\u001b[0;32m     50\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     51\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mcuda\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;31m#put the model on the GPU memory\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 52\u001b[1;33m       \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     53\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     54\u001b[0m   \u001b[1;32mdef\u001b[0m \u001b[0minit_weights\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mlayer\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;31m#gaussian init for the conv layers\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36mcuda\u001b[1;34m(self, device)\u001b[0m\n\u001b[0;32m    489\u001b[0m             \u001b[0mModule\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    490\u001b[0m         \"\"\"\n\u001b[1;32m--> 491\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;32mlambda\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    492\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    493\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mxpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mT\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdevice\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mOptional\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mUnion\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mint\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdevice\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m->\u001b[0m \u001b[0mT\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_apply\u001b[1;34m(self, fn)\u001b[0m\n\u001b[0;32m    385\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    386\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mchildren\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 387\u001b[1;33m             \u001b[0mmodule\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    388\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    389\u001b[0m         \u001b[1;32mdef\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtensor_applied\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_apply\u001b[1;34m(self, fn)\u001b[0m\n\u001b[0;32m    385\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    386\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mchildren\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 387\u001b[1;33m             \u001b[0mmodule\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfn\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    388\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    389\u001b[0m         \u001b[1;32mdef\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtensor_applied\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_apply\u001b[1;34m(self, fn)\u001b[0m\n\u001b[0;32m    407\u001b[0m                 \u001b[1;31m# `with torch.no_grad():`\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    408\u001b[0m                 \u001b[1;32mwith\u001b[0m \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mno_grad\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 409\u001b[1;33m                     \u001b[0mparam_applied\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparam\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    410\u001b[0m                 \u001b[0mshould_use_set_data\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcompute_should_use_set_data\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mparam\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mparam_applied\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    411\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[0mshould_use_set_data\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m<lambda>\u001b[1;34m(t)\u001b[0m\n\u001b[0;32m    489\u001b[0m             \u001b[0mModule\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    490\u001b[0m         \"\"\"\n\u001b[1;32m--> 491\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_apply\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;32mlambda\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mt\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    492\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    493\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mxpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mT\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdevice\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mOptional\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mUnion\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mint\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdevice\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m->\u001b[0m \u001b[0mT\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\cuda\\__init__.py\u001b[0m in \u001b[0;36m_lazy_init\u001b[1;34m()\u001b[0m\n\u001b[0;32m    162\u001b[0m                 \"multiprocessing, you must use the 'spawn' start method\")\n\u001b[0;32m    163\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_C\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'_cuda_getDeviceCount'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 164\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mAssertionError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Torch not compiled with CUDA enabled\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    165\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0m_cudart\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    166\u001b[0m             raise AssertionError(\n",
      "\u001b[1;31mAssertionError\u001b[0m: Torch not compiled with CUDA enabled"
     ]
    }
   ],
   "source": [
    "#[10]\n",
    "#==================TEST===============================\n",
    "#we consider the first point cloud from the training set\n",
    "tile, gt = tile_loader(0)\n",
    "segnet = SegNet(3,[16,16,32,32,64,32,32,16,32,16],6)\n",
    "print(segnet)\n",
    "print('Total number of parameters: {}'.format(sum([p.numel() for p in segnet.parameters()])))\n",
    "pred = segnet(tile[None,:,:,:]) #the None indicate a batch size of 1\n",
    "print(pred.shape)\n",
    "assert(pred.shape == torch.Size([1,6,64,64]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "id": "d60f58a1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SegNet(\n",
      "  (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  (unpool): MaxUnpool2d(kernel_size=(2, 2), stride=(2, 2), padding=(0, 0))\n",
      "  (c0): Sequential(\n",
      "    (0): Conv2d(4, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c1): Sequential(\n",
      "    (0): Conv2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c2): Sequential(\n",
      "    (0): Conv2d(16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c3): Sequential(\n",
      "    (0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c4): Sequential(\n",
      "    (0): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c5): Sequential(\n",
      "    (0): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c6): Sequential(\n",
      "    (0): Conv2d(64, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c7): Sequential(\n",
      "    (0): Conv2d(32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c8): Sequential(\n",
      "    (0): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c9): Sequential(\n",
      "    (0): Conv2d(32, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), padding_mode=reflect)\n",
      "    (1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (2): ReLU(inplace=True)\n",
      "  )\n",
      "  (c10): Conv2d(16, 6, kernel_size=(1, 1), stride=(1, 1), padding_mode=reflect)\n",
      ")\n",
      "Total number of parameters: 91398\n"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "Given groups=1, weight of size [16, 4, 3, 3], expected input[1, 3, 66, 66] to have 4 channels, but got 3 channels instead",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-118-82b9b26d47de>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      6\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msegnet\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      7\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'Total number of parameters: {}'\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msum\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnumel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mp\u001b[0m \u001b[1;32min\u001b[0m \u001b[0msegnet\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 8\u001b[1;33m \u001b[0mpred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0msegnet\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m#the None indicate a batch size of 1\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      9\u001b[0m \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpred\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshape\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     10\u001b[0m \u001b[1;32massert\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpred\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mshape\u001b[0m \u001b[1;33m==\u001b[0m \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mSize\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m6\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m256\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m256\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[1;34m(self, *input, **kwargs)\u001b[0m\n\u001b[0;32m    887\u001b[0m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    888\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 889\u001b[1;33m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    890\u001b[0m         for hook in itertools.chain(\n\u001b[0;32m    891\u001b[0m                 \u001b[0m_global_forward_hooks\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-116-2c4ea68d9187>\u001b[0m in \u001b[0;36mforward\u001b[1;34m(self, input)\u001b[0m\n\u001b[0;32m     63\u001b[0m     \u001b[1;31m#encoder\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     64\u001b[0m     \u001b[1;31m#level a\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 65\u001b[1;33m     \u001b[0mx1\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mc1\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mc0\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     66\u001b[0m     \u001b[0mx2\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mindices_a_b\u001b[0m \u001b[1;33m=\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmaxpool\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     67\u001b[0m     \u001b[1;31m#level b\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[1;34m(self, *input, **kwargs)\u001b[0m\n\u001b[0;32m    887\u001b[0m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    888\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 889\u001b[1;33m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    890\u001b[0m         for hook in itertools.chain(\n\u001b[0;32m    891\u001b[0m                 \u001b[0m_global_forward_hooks\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\container.py\u001b[0m in \u001b[0;36mforward\u001b[1;34m(self, input)\u001b[0m\n\u001b[0;32m    117\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    118\u001b[0m         \u001b[1;32mfor\u001b[0m \u001b[0mmodule\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 119\u001b[1;33m             \u001b[0minput\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodule\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    120\u001b[0m         \u001b[1;32mreturn\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    121\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[1;34m(self, *input, **kwargs)\u001b[0m\n\u001b[0;32m    887\u001b[0m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    888\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 889\u001b[1;33m             \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    890\u001b[0m         for hook in itertools.chain(\n\u001b[0;32m    891\u001b[0m                 \u001b[0m_global_forward_hooks\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\conv.py\u001b[0m in \u001b[0;36mforward\u001b[1;34m(self, input)\u001b[0m\n\u001b[0;32m    397\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    398\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 399\u001b[1;33m         \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_conv_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbias\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    400\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    401\u001b[0m \u001b[1;32mclass\u001b[0m \u001b[0mConv3d\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0m_ConvNd\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\nn\\modules\\conv.py\u001b[0m in \u001b[0;36m_conv_forward\u001b[1;34m(self, input, weight, bias)\u001b[0m\n\u001b[0;32m    392\u001b[0m             return F.conv2d(F.pad(input, self._reversed_padding_repeated_twice, mode=self.padding_mode),\n\u001b[0;32m    393\u001b[0m                             \u001b[0mweight\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbias\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstride\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 394\u001b[1;33m                             _pair(0), self.dilation, self.groups)\n\u001b[0m\u001b[0;32m    395\u001b[0m         return F.conv2d(input, weight, bias, self.stride,\n\u001b[0;32m    396\u001b[0m                         self.padding, self.dilation, self.groups)\n",
      "\u001b[1;31mRuntimeError\u001b[0m: Given groups=1, weight of size [16, 4, 3, 3], expected input[1, 3, 66, 66] to have 4 channels, but got 3 channels instead"
     ]
    }
   ],
   "source": [
    "#[10]\n",
    "#==================TEST===============================\n",
    "#we consider the first point cloud from the training set\n",
    "tile, gt = tile_loader(0)\n",
    "segnet = SegNet(4,[16,16,32,32,64,32,32,16,32,16],6)\n",
    "print(segnet)\n",
    "print('Total number of parameters: {}'.format(sum([p.numel() for p in segnet.parameters()])))\n",
    "pred = segnet(tile[None,:,:,:]) #the None indicate a batch size of 1\n",
    "print(pred.shape)\n",
    "assert(pred.shape == torch.Size([1,6,64,64]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "id": "78ecb290",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of parameters: 91254\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "92ebcb1230d44889bbd3290cfc07d29d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 4.90%  |  infested water : 3.75%\n",
      "Epoch   0 -> Train Overall Accuracy: 5.00% Train mIoU : 1.44% Train Loss: 2.7093\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "63da0cc74f6f48c5ac4a4e3e01cabfa6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 7.83%  |  infested water : 36.02%\n",
      "Epoch   1 -> Train Overall Accuracy: 9.81% Train mIoU : 7.31% Train Loss: 2.3102\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c715245215ae41fd80706ae5caafc910",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 14.81%  |  infested water : 27.99%\n",
      "Epoch   2 -> Train Overall Accuracy: 15.92% Train mIoU : 7.13% Train Loss: 2.0176\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "338f71afd9574ae9bb39d4f18e1c2417",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 23.07%  |  infested water : 26.03%\n",
      "Epoch   3 -> Train Overall Accuracy: 23.54% Train mIoU : 8.18% Train Loss: 1.7811\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9fb7f3dca39449589ded3f721fbb3407",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 27.09%  |  infested water : 36.45%\n",
      "Epoch   4 -> Train Overall Accuracy: 28.60% Train mIoU : 10.59% Train Loss: 1.6040\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0b39780091b5473997aa3e8d76411c3e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 40.97%  |  infested water : 43.61%\n",
      "Epoch   5 -> Train Overall Accuracy: 42.15% Train mIoU : 14.10% Train Loss: 1.4627\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "547c3bca64ca43bfb843c578ddd20f04",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[104m\n",
      "clean water : 0.04%  |  infested water : 5.25%\n",
      "Test Overall Accuracy: 3.58% Test mIoU : 0.88%  Test Loss: 134.1993\u001b[0m\n"
     ]
    },
    {
     "ename": "AssertionError",
     "evalue": "Torch not compiled with CUDA enabled",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-132-b60d3ff766d4>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     10\u001b[0m \u001b[0margs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlr\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m5e-3\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     11\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 12\u001b[1;33m \u001b[0mtrained_model\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrain_full\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-63-6490caa954e7>\u001b[0m in \u001b[0;36mtrain_full\u001b[1;34m(args)\u001b[0m\n\u001b[0;32m     99\u001b[0m       \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mTESTCOLOR\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    100\u001b[0m       \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'Test Overall Accuracy: %3.2f%% Test mIoU : %3.2f%%  Test Loss: %1.4f'\u001b[0m \u001b[1;33m%\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mcm_test\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moverall_accuracy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcm_test\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mclass_IoU\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mloss_test\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mNORMALCOLOR\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 101\u001b[1;33m       \u001b[0mviewer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mn_shown\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcategory\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'cigpe'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0muse_mask\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    102\u001b[0m   \u001b[1;32mreturn\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-124-ec42920f6a40>\u001b[0m in \u001b[0;36mviewer\u001b[1;34m(n_shown, category, train, model, use_mask)\u001b[0m\n\u001b[0;32m     44\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0msubplot_index\u001b[0m \u001b[1;33m<=\u001b[0m \u001b[0mn_category\u001b[0m \u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     45\u001b[0m         \u001b[0max\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mset\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtitle\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'Prediction'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 46\u001b[1;33m       \u001b[0mpred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msqueeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     47\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0muse_mask\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     48\u001b[0m         \u001b[0mpred\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mgt\u001b[0m\u001b[1;33m==\u001b[0m\u001b[1;36m99\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m99\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\cuda\\__init__.py\u001b[0m in \u001b[0;36m_lazy_init\u001b[1;34m()\u001b[0m\n\u001b[0;32m    162\u001b[0m                 \"multiprocessing, you must use the 'spawn' start method\")\n\u001b[0;32m    163\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_C\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'_cuda_getDeviceCount'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 164\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mAssertionError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Torch not compiled with CUDA enabled\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    165\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0m_cudart\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    166\u001b[0m             raise AssertionError(\n",
      "\u001b[1;31mAssertionError\u001b[0m: Torch not compiled with CUDA enabled"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x360 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#[12]\n",
    "args = mock.Mock() #stores the parameters\n",
    "args.n_epoch = 50\n",
    "args.n_epoch_test = int(5) #periodicity of evaluation on test set\n",
    "args.batch_size = 16\n",
    "args.n_class = 6\n",
    "args.n_channel = 3\n",
    "args.conv_width = [16,16,32,32,64,32,32,16,32,16]\n",
    "args.cuda = 1\n",
    "args.lr = 5e-3\n",
    "\n",
    "trained_model = train_full(args)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "9f1360b3",
   "metadata": {},
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "tile_loader() got an unexpected keyword argument 'cuda'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-73-fbcdfcf2e578>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mtile\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgt\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtile_loader\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m6\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrain\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcuda\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0mpred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrained_model\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msqueeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msave\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'pred.npy'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmoveaxis\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpred\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdetach\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnumpy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m-\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0mfiles\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdownload\u001b[0m\u001b[1;33m(\u001b[0m \u001b[1;34m\"pred.npy\"\u001b[0m \u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mview_rgb\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0max\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mTypeError\u001b[0m: tile_loader() got an unexpected keyword argument 'cuda'"
     ]
    }
   ],
   "source": [
    "tile, gt = tile_loader(6, train = train, cuda=0)\n",
    "pred = trained_model(tile[None,:,:,:].cuda()).cpu().squeeze()\n",
    "np.save('pred.npy', np.moveaxis(pred.detach().numpy(), 0, -1))\n",
    "files.download( \"pred.npy\" )\n",
    "view_rgb(tile, ax = None)\n",
    "view_labels(pred.argmax(0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "eb0c93b3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of parameters: 91254\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "05f8bf2a7ab14ba6a8d4e028e112790d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 0.44%  |  infested water : 7.42%\n",
      "Epoch   0 -> Train Overall Accuracy: 1.94% Train mIoU : 1.31% Train Loss: 5.0535\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fe299b311d064e0e9ef377727e0d3b16",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 0.44%  |  infested water : 19.30%\n",
      "Epoch   1 -> Train Overall Accuracy: 3.48% Train mIoU : 3.29% Train Loss: 4.0691\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2fc5cf8b7a394a7ea2128a892357e70b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 1.34%  |  infested water : 20.10%\n",
      "Epoch   2 -> Train Overall Accuracy: 3.77% Train mIoU : 3.57% Train Loss: 3.0860\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "414f55da991847379e4d7d5e24ef3d6f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 8.62%  |  infested water : 19.75%\n",
      "Epoch   3 -> Train Overall Accuracy: 10.38% Train mIoU : 4.73% Train Loss: 2.5628\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1b5874f08f1a4f809d3a45dd4273b0e1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 18.01%  |  infested water : 27.56%\n",
      "Epoch   4 -> Train Overall Accuracy: 19.54% Train mIoU : 7.59% Train Loss: 2.2734\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "38ee81f0ad7f42a1be10766922f00645",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[100m\n",
      "clean water : 22.82%  |  infested water : 41.96%\n",
      "Epoch   5 -> Train Overall Accuracy: 25.06% Train mIoU : 10.80% Train Loss: 1.9931\u001b[0m\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8879517ab87d49c29a35015be5f71e4a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|                                                                                                         …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[104m\n",
      "clean water : 0.00%  |  infested water : 5.46%\n",
      "Test Overall Accuracy: 4.98% Test mIoU : 1.82%  Test Loss: 333.6662\u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\vaak\\Anaconda3\\envs\\snappy\\lib\\site-packages\\ipykernel_launcher.py:18: RuntimeWarning:\n",
      "\n",
      "invalid value encountered in true_divide\n",
      "\n"
     ]
    },
    {
     "ename": "AssertionError",
     "evalue": "Torch not compiled with CUDA enabled",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mAssertionError\u001b[0m                            Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-82-b23f07723fdd>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mtrained_model\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrain_full\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[0mpred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrained_model\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msqueeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-63-6490caa954e7>\u001b[0m in \u001b[0;36mtrain_full\u001b[1;34m(args)\u001b[0m\n\u001b[0;32m     99\u001b[0m       \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mTESTCOLOR\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    100\u001b[0m       \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'Test Overall Accuracy: %3.2f%% Test mIoU : %3.2f%%  Test Loss: %1.4f'\u001b[0m \u001b[1;33m%\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mcm_test\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moverall_accuracy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcm_test\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mclass_IoU\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mloss_test\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m+\u001b[0m \u001b[0mNORMALCOLOR\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 101\u001b[1;33m       \u001b[0mviewer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mn_shown\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcategory\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'cigpe'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0muse_mask\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    102\u001b[0m   \u001b[1;32mreturn\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-39-1928740b3b17>\u001b[0m in \u001b[0;36mviewer\u001b[1;34m(n_shown, category, train, model, use_mask)\u001b[0m\n\u001b[0;32m     44\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0msubplot_index\u001b[0m \u001b[1;33m<=\u001b[0m \u001b[0mn_category\u001b[0m \u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     45\u001b[0m         \u001b[0max\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mset\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtitle\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'Prediction'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 46\u001b[1;33m       \u001b[0mpred\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtile\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcpu\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0margmax\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msqueeze\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     47\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0muse_mask\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     48\u001b[0m         \u001b[0mpred\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mgt\u001b[0m\u001b[1;33m==\u001b[0m\u001b[1;36m99\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m99\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\snappy\\lib\\site-packages\\torch\\cuda\\__init__.py\u001b[0m in \u001b[0;36m_lazy_init\u001b[1;34m()\u001b[0m\n\u001b[0;32m    162\u001b[0m                 \"multiprocessing, you must use the 'spawn' start method\")\n\u001b[0;32m    163\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mhasattr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_C\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'_cuda_getDeviceCount'\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 164\u001b[1;33m             \u001b[1;32mraise\u001b[0m \u001b[0mAssertionError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"Torch not compiled with CUDA enabled\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    165\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0m_cudart\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    166\u001b[0m             raise AssertionError(\n",
      "\u001b[1;31mAssertionError\u001b[0m: Torch not compiled with CUDA enabled"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x360 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "trained_model = train_full(args)\n",
    "\n",
    "pred = trained_model(tile[None,:,:,:].cuda()).cpu().squeeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "8955e2ad",
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'trained_model' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-74-8cdafebad1d8>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mviewer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mn_shown\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m5\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcategory\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'cigpe'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrained_model\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtrain\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0muse_mask\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m: name 'trained_model' is not defined"
     ]
    }
   ],
   "source": [
    "viewer(n_shown=5, category = 'cigpe', model = trained_model, train = False, use_mask = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bc002678",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}