dataprep.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import glob\n",
    "from tqdm import tqdm_notebook\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200\n"
     ]
    }
   ],
   "source": [
    "labels = os.listdir(\"tiny-imagenet-200/train\")\n",
    "print(len(labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "unseen_test_labels = ['n02814533', 'n02236044', 'n04328186', 'n02423022', 'n06596364', 'n02892201', 'n07749582', 'n01944390', 'n04532106', 'n02129165']\n",
    "labels = [y for y in labels if y not in unseen_test_labels]\n",
    "labels = random.sample(labels, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "sorted_labels = sorted(labels)\n",
    "labels_dict = {\"label2id\": {}, \"id2label\": {}}\n",
    "for i, label in enumerate(sorted_labels):\n",
    "    labels_dict[\"label2id\"][label] = i\n",
    "    labels_dict[\"id2label\"][i] = label\n",
    "\n",
    "import json\n",
    "json.dump(labels_dict, open(\"labels_dict.json\", \"w\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = []\n",
    "for y in labels:\n",
    "    images = os.listdir(os.path.join(\"tiny-imagenet-200\", \"train\", y, \"images\"))\n",
    "    for x in images:\n",
    "        path = f\"train/{y}/images/{x}\"\n",
    "        train_data.append({\n",
    "            \"image_path\": path,\n",
    "            \"label\": y\n",
    "        })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image_path</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>train/n02769748/images/n02769748_0.JPEG</td>\n",
       "      <td>n02769748</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train/n02769748/images/n02769748_1.JPEG</td>\n",
       "      <td>n02769748</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>train/n02769748/images/n02769748_10.JPEG</td>\n",
       "      <td>n02769748</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>train/n02769748/images/n02769748_100.JPEG</td>\n",
       "      <td>n02769748</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>train/n02769748/images/n02769748_101.JPEG</td>\n",
       "      <td>n02769748</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24995</th>\n",
       "      <td>train/n03444034/images/n03444034_95.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24996</th>\n",
       "      <td>train/n03444034/images/n03444034_96.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24997</th>\n",
       "      <td>train/n03444034/images/n03444034_97.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24998</th>\n",
       "      <td>train/n03444034/images/n03444034_98.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24999</th>\n",
       "      <td>train/n03444034/images/n03444034_99.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>25000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                      image_path      label\n",
       "0        train/n02769748/images/n02769748_0.JPEG  n02769748\n",
       "1        train/n02769748/images/n02769748_1.JPEG  n02769748\n",
       "2       train/n02769748/images/n02769748_10.JPEG  n02769748\n",
       "3      train/n02769748/images/n02769748_100.JPEG  n02769748\n",
       "4      train/n02769748/images/n02769748_101.JPEG  n02769748\n",
       "...                                          ...        ...\n",
       "24995   train/n03444034/images/n03444034_95.JPEG  n03444034\n",
       "24996   train/n03444034/images/n03444034_96.JPEG  n03444034\n",
       "24997   train/n03444034/images/n03444034_97.JPEG  n03444034\n",
       "24998   train/n03444034/images/n03444034_98.JPEG  n03444034\n",
       "24999   train/n03444034/images/n03444034_99.JPEG  n03444034\n",
       "\n",
       "[25000 rows x 2 columns]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = pd.DataFrame(train_data)\n",
    "train_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "n02769748    500\n",
       "n02950826    500\n",
       "n02963159    500\n",
       "n02099712    500\n",
       "n04311004    500\n",
       "n03126707    500\n",
       "n02909870    500\n",
       "n04501370    500\n",
       "n03649909    500\n",
       "n01945685    500\n",
       "n01917289    500\n",
       "n04146614    500\n",
       "n02233338    500\n",
       "n07579787    500\n",
       "n03404251    500\n",
       "n09332890    500\n",
       "n04486054    500\n",
       "n07768694    500\n",
       "n02268443    500\n",
       "n04560804    500\n",
       "n03976657    500\n",
       "n03930313    500\n",
       "n03983396    500\n",
       "n03355925    500\n",
       "n07615774    500\n",
       "n04399382    500\n",
       "n02509815    500\n",
       "n03584254    500\n",
       "n04507155    500\n",
       "n02883205    500\n",
       "n03447447    500\n",
       "n03617480    500\n",
       "n04356056    500\n",
       "n02190166    500\n",
       "n02917067    500\n",
       "n02106662    500\n",
       "n04149813    500\n",
       "n01983481    500\n",
       "n04254777    500\n",
       "n02403003    500\n",
       "n04371430    500\n",
       "n02165456    500\n",
       "n07734744    500\n",
       "n07873807    500\n",
       "n07871810    500\n",
       "n03814639    500\n",
       "n02437312    500\n",
       "n03662601    500\n",
       "n04023962    500\n",
       "n03444034    500\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df.to_csv('train.csv', index=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Val Test Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "seen_test_indices = [5, 6, 10, 11, 12, 15, 16, 18, 22, 25, 27, 29, 32, 33, 36, 37, 43, 44, 48, 49]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"tiny-imagenet-200/val/val_annotations.txt\", \"r\") as f:\n",
    "    lines = f.readlines()\n",
    "\n",
    "val_samples = {}\n",
    "for line in lines:\n",
    "    tmp = line.strip().split('\\t')\n",
    "    path = f\"val/images/{tmp[0]}\"\n",
    "    label = tmp[1]\n",
    "    if label in unseen_test_labels: continue\n",
    "    if label not in labels: continue\n",
    "\n",
    "    if label in val_samples: val_samples[label].append(path)\n",
    "    else: val_samples[label] = [path]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_data = []\n",
    "seen_test_samples = {}\n",
    "\n",
    "for y, x in val_samples.items():\n",
    "    for i, path in enumerate(x):\n",
    "        if i in seen_test_indices:\n",
    "            if y in seen_test_samples: seen_test_samples[y].append(path)\n",
    "            else: seen_test_samples[y] = [path]\n",
    "        else: val_data.append({\n",
    "            \"image_path\": path,\n",
    "            \"label\": y\n",
    "        })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image_path</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>val/images/val_0.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>val/images/val_284.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val/images/val_355.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>val/images/val_505.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>val/images/val_1007.JPEG</td>\n",
       "      <td>n03444034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1495</th>\n",
       "      <td>val/images/val_8414.JPEG</td>\n",
       "      <td>n03976657</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1496</th>\n",
       "      <td>val/images/val_9137.JPEG</td>\n",
       "      <td>n03976657</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1497</th>\n",
       "      <td>val/images/val_9681.JPEG</td>\n",
       "      <td>n03976657</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1498</th>\n",
       "      <td>val/images/val_9731.JPEG</td>\n",
       "      <td>n03976657</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1499</th>\n",
       "      <td>val/images/val_9806.JPEG</td>\n",
       "      <td>n03976657</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1500 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    image_path      label\n",
       "0        val/images/val_0.JPEG  n03444034\n",
       "1      val/images/val_284.JPEG  n03444034\n",
       "2      val/images/val_355.JPEG  n03444034\n",
       "3      val/images/val_505.JPEG  n03444034\n",
       "4     val/images/val_1007.JPEG  n03444034\n",
       "...                        ...        ...\n",
       "1495  val/images/val_8414.JPEG  n03976657\n",
       "1496  val/images/val_9137.JPEG  n03976657\n",
       "1497  val/images/val_9681.JPEG  n03976657\n",
       "1498  val/images/val_9731.JPEG  n03976657\n",
       "1499  val/images/val_9806.JPEG  n03976657\n",
       "\n",
       "[1500 rows x 2 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_df = pd.DataFrame(val_data)\n",
    "val_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "n03444034    30\n",
       "n02268443    30\n",
       "n02403003    30\n",
       "n03662601    30\n",
       "n04254777    30\n",
       "n02233338    30\n",
       "n02509815    30\n",
       "n02883205    30\n",
       "n03447447    30\n",
       "n04356056    30\n",
       "n07873807    30\n",
       "n02917067    30\n",
       "n03983396    30\n",
       "n04399382    30\n",
       "n03649909    30\n",
       "n03404251    30\n",
       "n04507155    30\n",
       "n04149813    30\n",
       "n02099712    30\n",
       "n03617480    30\n",
       "n09332890    30\n",
       "n02963159    30\n",
       "n02106662    30\n",
       "n04501370    30\n",
       "n04311004    30\n",
       "n07768694    30\n",
       "n07615774    30\n",
       "n02950826    30\n",
       "n04146614    30\n",
       "n04371430    30\n",
       "n04486054    30\n",
       "n03814639    30\n",
       "n01983481    30\n",
       "n03930313    30\n",
       "n03584254    30\n",
       "n02909870    30\n",
       "n01917289    30\n",
       "n07579787    30\n",
       "n03355925    30\n",
       "n03126707    30\n",
       "n02165456    30\n",
       "n01945685    30\n",
       "n02437312    30\n",
       "n02769748    30\n",
       "n04560804    30\n",
       "n04023962    30\n",
       "n02190166    30\n",
       "n07871810    30\n",
       "n07734744    30\n",
       "n03976657    30\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_df.label.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_df.to_csv('val.csv', index=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create Seen Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "seen_seen_test_data = []\n",
    "\n",
    "for y, x in seen_test_samples.items():\n",
    "    random.shuffle(x)\n",
    "    for i in range(10):\n",
    "        seen_seen_test_data.append({\n",
    "            'image1': x[i],\n",
    "            'image2': x[len(x)-i-1],\n",
    "            'label': 1\n",
    "        })\n",
    "    \n",
    "    random.shuffle(x)\n",
    "    for i in range(10):\n",
    "        neg_id = random.sample([elem for i, elem in enumerate(seen_test_samples.keys()) if i != y], 1)[0]\n",
    "        neg_sample = random.sample(seen_test_samples[neg_id], 1)[0]\n",
    "        seen_seen_test_data.append({\n",
    "            'image1': x[i],\n",
    "            'image2': neg_sample,\n",
    "            'label': 0\n",
    "        })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image1</th>\n",
       "      <th>image2</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>val/images/val_5453.JPEG</td>\n",
       "      <td>val/images/val_9939.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>val/images/val_8190.JPEG</td>\n",
       "      <td>val/images/val_9382.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val/images/val_7470.JPEG</td>\n",
       "      <td>val/images/val_3433.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>val/images/val_1008.JPEG</td>\n",
       "      <td>val/images/val_6988.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>val/images/val_2134.JPEG</td>\n",
       "      <td>val/images/val_4839.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>995</th>\n",
       "      <td>val/images/val_3626.JPEG</td>\n",
       "      <td>val/images/val_6368.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>996</th>\n",
       "      <td>val/images/val_3078.JPEG</td>\n",
       "      <td>val/images/val_3341.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>997</th>\n",
       "      <td>val/images/val_7299.JPEG</td>\n",
       "      <td>val/images/val_5041.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>998</th>\n",
       "      <td>val/images/val_3635.JPEG</td>\n",
       "      <td>val/images/val_2072.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999</th>\n",
       "      <td>val/images/val_1740.JPEG</td>\n",
       "      <td>val/images/val_9299.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1000 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       image1                    image2  label\n",
       "0    val/images/val_5453.JPEG  val/images/val_9939.JPEG      1\n",
       "1    val/images/val_8190.JPEG  val/images/val_9382.JPEG      1\n",
       "2    val/images/val_7470.JPEG  val/images/val_3433.JPEG      1\n",
       "3    val/images/val_1008.JPEG  val/images/val_6988.JPEG      1\n",
       "4    val/images/val_2134.JPEG  val/images/val_4839.JPEG      1\n",
       "..                        ...                       ...    ...\n",
       "995  val/images/val_3626.JPEG  val/images/val_6368.JPEG      0\n",
       "996  val/images/val_3078.JPEG  val/images/val_3341.JPEG      0\n",
       "997  val/images/val_7299.JPEG  val/images/val_5041.JPEG      0\n",
       "998  val/images/val_3635.JPEG  val/images/val_2072.JPEG      0\n",
       "999  val/images/val_1740.JPEG  val/images/val_9299.JPEG      0\n",
       "\n",
       "[1000 rows x 3 columns]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seen_seen_test_df = pd.DataFrame(seen_seen_test_data)\n",
    "seen_seen_test_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "seen_seen_test_df.to_csv('seen_seen_test.csv', index=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create Unseen Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"tiny-imagenet-200/val/val_annotations.txt\", \"r\") as f:\n",
    "    lines = f.readlines()\n",
    "\n",
    "unseen_test_samples = {}\n",
    "for line in lines:\n",
    "    tmp = line.strip().split('\\t')\n",
    "    path = os.path.join(\"val\", \"images\", tmp[0]).replace('\\\\', '/')\n",
    "    label = tmp[1]\n",
    "    if label in unseen_test_labels:\n",
    "        if label in unseen_test_samples: unseen_test_samples[label].append(path)\n",
    "        else: unseen_test_samples[label] = [path]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "unseen_unseen_test_data = []\n",
    "\n",
    "for y, x in unseen_test_samples.items():\n",
    "    for i in range(10):\n",
    "        aid = random.randint(0, len(x)-1)\n",
    "        anchor = x[aid]\n",
    "        pos_sample = random.sample([elem for i, elem in enumerate(x) if i != aid], 1)\n",
    "        unseen_unseen_test_data.append({\n",
    "            'image1': anchor,\n",
    "            'image2': pos_sample[0],\n",
    "            'label': 1\n",
    "        })\n",
    "\n",
    "        neg_id = random.sample([elem for i, elem in enumerate(unseen_test_samples.keys()) if i != y], 1)\n",
    "        neg_sample = random.sample(unseen_test_samples[neg_id[0]], 1)\n",
    "        unseen_unseen_test_data.append({\n",
    "            'image1': anchor,\n",
    "            'image2': neg_sample[0],\n",
    "            'label': 0\n",
    "        })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image1</th>\n",
       "      <th>image2</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>val/images/val_1671.JPEG</td>\n",
       "      <td>val/images/val_84.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>val/images/val_1671.JPEG</td>\n",
       "      <td>val/images/val_4914.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val/images/val_4533.JPEG</td>\n",
       "      <td>val/images/val_4460.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>val/images/val_4533.JPEG</td>\n",
       "      <td>val/images/val_6696.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>val/images/val_2066.JPEG</td>\n",
       "      <td>val/images/val_2234.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>195</th>\n",
       "      <td>val/images/val_1057.JPEG</td>\n",
       "      <td>val/images/val_7582.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>196</th>\n",
       "      <td>val/images/val_6257.JPEG</td>\n",
       "      <td>val/images/val_3925.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>197</th>\n",
       "      <td>val/images/val_6257.JPEG</td>\n",
       "      <td>val/images/val_567.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>198</th>\n",
       "      <td>val/images/val_6198.JPEG</td>\n",
       "      <td>val/images/val_8155.JPEG</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>199</th>\n",
       "      <td>val/images/val_6198.JPEG</td>\n",
       "      <td>val/images/val_8658.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>200 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       image1                    image2  label\n",
       "0    val/images/val_1671.JPEG    val/images/val_84.JPEG      1\n",
       "1    val/images/val_1671.JPEG  val/images/val_4914.JPEG      0\n",
       "2    val/images/val_4533.JPEG  val/images/val_4460.JPEG      1\n",
       "3    val/images/val_4533.JPEG  val/images/val_6696.JPEG      0\n",
       "4    val/images/val_2066.JPEG  val/images/val_2234.JPEG      1\n",
       "..                        ...                       ...    ...\n",
       "195  val/images/val_1057.JPEG  val/images/val_7582.JPEG      0\n",
       "196  val/images/val_6257.JPEG  val/images/val_3925.JPEG      1\n",
       "197  val/images/val_6257.JPEG   val/images/val_567.JPEG      0\n",
       "198  val/images/val_6198.JPEG  val/images/val_8155.JPEG      1\n",
       "199  val/images/val_6198.JPEG  val/images/val_8658.JPEG      0\n",
       "\n",
       "[200 rows x 3 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unseen_unseen_test_df = pd.DataFrame(unseen_unseen_test_data)\n",
    "unseen_unseen_test_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "unseen_unseen_test_df.to_csv('unseen_unseen_test.csv', index=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Seen Unseen Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_labels = val_df.label.unique()\n",
    "seen_test_labels = random.sample(list(val_labels), 10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "random.shuffle(unseen_test_labels)\n",
    "seen_unseen_test_data = []\n",
    "for seen_label, unseen_label in zip(seen_test_labels, unseen_test_labels):\n",
    "    seen_paths = random.sample(seen_test_samples[seen_label], 20)\n",
    "    unseen_paths = random.sample(unseen_test_samples[unseen_label], 20)\n",
    "    for sp, up in zip(seen_paths, unseen_paths):\n",
    "        seen_unseen_test_data.append({\n",
    "            'image1': sp,\n",
    "            'image2': up,\n",
    "            'label': 0\n",
    "        })"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>image1</th>\n",
       "      <th>image2</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>val/images/val_8508.JPEG</td>\n",
       "      <td>val/images/val_3925.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>val/images/val_1528.JPEG</td>\n",
       "      <td>val/images/val_1475.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>val/images/val_4799.JPEG</td>\n",
       "      <td>val/images/val_1232.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>val/images/val_1516.JPEG</td>\n",
       "      <td>val/images/val_6593.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>val/images/val_3859.JPEG</td>\n",
       "      <td>val/images/val_1850.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>195</th>\n",
       "      <td>val/images/val_7043.JPEG</td>\n",
       "      <td>val/images/val_1763.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>196</th>\n",
       "      <td>val/images/val_3473.JPEG</td>\n",
       "      <td>val/images/val_6263.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>197</th>\n",
       "      <td>val/images/val_5237.JPEG</td>\n",
       "      <td>val/images/val_4266.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>198</th>\n",
       "      <td>val/images/val_6988.JPEG</td>\n",
       "      <td>val/images/val_1378.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>199</th>\n",
       "      <td>val/images/val_2507.JPEG</td>\n",
       "      <td>val/images/val_3007.JPEG</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>200 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                       image1                    image2  label\n",
       "0    val/images/val_8508.JPEG  val/images/val_3925.JPEG      0\n",
       "1    val/images/val_1528.JPEG  val/images/val_1475.JPEG      0\n",
       "2    val/images/val_4799.JPEG  val/images/val_1232.JPEG      0\n",
       "3    val/images/val_1516.JPEG  val/images/val_6593.JPEG      0\n",
       "4    val/images/val_3859.JPEG  val/images/val_1850.JPEG      0\n",
       "..                        ...                       ...    ...\n",
       "195  val/images/val_7043.JPEG  val/images/val_1763.JPEG      0\n",
       "196  val/images/val_3473.JPEG  val/images/val_6263.JPEG      0\n",
       "197  val/images/val_5237.JPEG  val/images/val_4266.JPEG      0\n",
       "198  val/images/val_6988.JPEG  val/images/val_1378.JPEG      0\n",
       "199  val/images/val_2507.JPEG  val/images/val_3007.JPEG      0\n",
       "\n",
       "[200 rows x 3 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seen_unseen_test_df = pd.DataFrame(seen_unseen_test_data)\n",
    "seen_unseen_test_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "seen_unseen_test_df.to_csv('seen_unseen_test.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "DLenv",
   "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": "3.8.13"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "6e27fdd685364325128668c401744e268eed4a605f1a6553ece99da66ad5edbc"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}