ANN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "c9efcbe4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import shutil\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torchvision import datasets, transforms\n",
    "from torch.autograd import Variable\n",
    "import os\n",
    "import pandas as pd\n",
    "from torchvision.io import read_image\n",
    "from torch.utils.data import Dataset\n",
    "import cv2\n",
    "from torch.utils.data import DataLoader\n",
    "import numpy as np\n",
    "import copy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e722de79",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/Users/harsh/ML/Project/project\r\n"
     ]
    }
   ],
   "source": [
    "!pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "6a9c45c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics\n",
    "def metric(y_test, y_pred):\n",
    "    matrix = metrics.confusion_matrix(y_test, y_pred)\n",
    "    accuracy = metrics.accuracy_score(y_test, y_pred)\n",
    "    precision = metrics.precision_score(y_test, y_pred)\n",
    "    recall = metrics.recall_score(y_test, y_pred)\n",
    "    f1 = metrics.f1_score(y_test, y_pred)\n",
    "    return matrix,accuracy,precision,recall,f1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "434bf49a",
   "metadata": {},
   "outputs": [],
   "source": [
    "def move_files(abs_dirname, train_dir, test_dir):\n",
    "    \"\"\"Move files into subdirectories.\"\"\"\n",
    "\n",
    "    files = [os.path.join(abs_dirname, f) for f in os.listdir(abs_dirname)]\n",
    "\n",
    "    i = 0\n",
    "    curr_subdir = None\n",
    "    files.sort()\n",
    "\n",
    "    for f in files:\n",
    "        # create new subdir if necessary\n",
    "        if i == 0:\n",
    "            curr_subdir = train_dir\n",
    "        \n",
    "        if i == 11025:\n",
    "            curr_subdir = test_dir\n",
    "\n",
    "        # move file to current dir\n",
    "        f_base = os.path.basename(f)\n",
    "        shutil.copy(f, os.path.join(curr_subdir, f_base))\n",
    "        i += 1\n",
    "        if(i%100 == 0):\n",
    "            print(i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "a8b8cd76",
   "metadata": {},
   "outputs": [],
   "source": [
    "abs_dirname = \"/Users/harsh/ML/Project/cell_images/Parasitized/\"\n",
    "train_dir = \"/Users/harsh/ML/Project/project/dataset/train/parasite\"\n",
    "test_dir = \"/Users/harsh/ML/Project/project/dataset/test/parasite\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "707c5f1e",
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100\n",
      "200\n",
      "300\n",
      "400\n",
      "500\n",
      "600\n",
      "700\n",
      "800\n",
      "900\n",
      "1000\n",
      "1100\n",
      "1200\n",
      "1300\n",
      "1400\n",
      "1500\n",
      "1600\n",
      "1700\n",
      "1800\n",
      "1900\n",
      "2000\n",
      "2100\n",
      "2200\n",
      "2300\n",
      "2400\n",
      "2500\n",
      "2600\n",
      "2700\n",
      "2800\n",
      "2900\n",
      "3000\n",
      "3100\n",
      "3200\n",
      "3300\n",
      "3400\n",
      "3500\n",
      "3600\n",
      "3700\n",
      "3800\n",
      "3900\n",
      "4000\n",
      "4100\n",
      "4200\n",
      "4300\n",
      "4400\n",
      "4500\n",
      "4600\n",
      "4700\n",
      "4800\n",
      "4900\n",
      "5000\n",
      "5100\n",
      "5200\n",
      "5300\n",
      "5400\n",
      "5500\n",
      "5600\n",
      "5700\n",
      "5800\n",
      "5900\n",
      "6000\n",
      "6100\n",
      "6200\n",
      "6300\n",
      "6400\n",
      "6500\n",
      "6600\n",
      "6700\n",
      "6800\n",
      "6900\n",
      "7000\n",
      "7100\n",
      "7200\n",
      "7300\n",
      "7400\n",
      "7500\n",
      "7600\n",
      "7700\n",
      "7800\n",
      "7900\n",
      "8000\n",
      "8100\n",
      "8200\n",
      "8300\n",
      "8400\n",
      "8500\n",
      "8600\n",
      "8700\n",
      "8800\n",
      "8900\n",
      "9000\n",
      "9100\n",
      "9200\n",
      "9300\n",
      "9400\n",
      "9500\n",
      "9600\n",
      "9700\n",
      "9800\n",
      "9900\n",
      "10000\n",
      "10100\n",
      "10200\n",
      "10300\n",
      "10400\n",
      "10500\n",
      "10600\n",
      "10700\n",
      "10800\n",
      "10900\n",
      "11000\n",
      "11100\n",
      "11200\n",
      "11300\n",
      "11400\n",
      "11500\n",
      "11600\n",
      "11700\n",
      "11800\n",
      "11900\n",
      "12000\n",
      "12100\n",
      "12200\n",
      "12300\n",
      "12400\n",
      "12500\n",
      "12600\n",
      "12700\n",
      "12800\n",
      "12900\n",
      "13000\n",
      "13100\n",
      "13200\n",
      "13300\n",
      "13400\n",
      "13500\n",
      "13600\n",
      "13700\n"
     ]
    }
   ],
   "source": [
    "move_files(abs_dirname, train_dir, test_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "02444d5d",
   "metadata": {},
   "outputs": [],
   "source": [
    "abs_dirname = \"/Users/harsh/ML/Project/cell_images/Uninfected/\"\n",
    "train_dir = \"/Users/harsh/ML/Project/project/dataset/train/uninfected\"\n",
    "test_dir = \"/Users/harsh/ML/Project/project/dataset/test/uninfected\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "4195a39a",
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100\n",
      "200\n",
      "300\n",
      "400\n",
      "500\n",
      "600\n",
      "700\n",
      "800\n",
      "900\n",
      "1000\n",
      "1100\n",
      "1200\n",
      "1300\n",
      "1400\n",
      "1500\n",
      "1600\n",
      "1700\n",
      "1800\n",
      "1900\n",
      "2000\n",
      "2100\n",
      "2200\n",
      "2300\n",
      "2400\n",
      "2500\n",
      "2600\n",
      "2700\n",
      "2800\n",
      "2900\n",
      "3000\n",
      "3100\n",
      "3200\n",
      "3300\n",
      "3400\n",
      "3500\n",
      "3600\n",
      "3700\n",
      "3800\n",
      "3900\n",
      "4000\n",
      "4100\n",
      "4200\n",
      "4300\n",
      "4400\n",
      "4500\n",
      "4600\n",
      "4700\n",
      "4800\n",
      "4900\n",
      "5000\n",
      "5100\n",
      "5200\n",
      "5300\n",
      "5400\n",
      "5500\n",
      "5600\n",
      "5700\n",
      "5800\n",
      "5900\n",
      "6000\n",
      "6100\n",
      "6200\n",
      "6300\n",
      "6400\n",
      "6500\n",
      "6600\n",
      "6700\n",
      "6800\n",
      "6900\n",
      "7000\n",
      "7100\n",
      "7200\n",
      "7300\n",
      "7400\n",
      "7500\n",
      "7600\n",
      "7700\n",
      "7800\n",
      "7900\n",
      "8000\n",
      "8100\n",
      "8200\n",
      "8300\n",
      "8400\n",
      "8500\n",
      "8600\n",
      "8700\n",
      "8800\n",
      "8900\n",
      "9000\n",
      "9100\n",
      "9200\n",
      "9300\n",
      "9400\n",
      "9500\n",
      "9600\n",
      "9700\n",
      "9800\n",
      "9900\n",
      "10000\n",
      "10100\n",
      "10200\n",
      "10300\n",
      "10400\n",
      "10500\n",
      "10600\n",
      "10700\n",
      "10800\n",
      "10900\n",
      "11000\n",
      "11100\n",
      "11200\n",
      "11300\n",
      "11400\n",
      "11500\n",
      "11600\n",
      "11700\n",
      "11800\n",
      "11900\n",
      "12000\n",
      "12100\n",
      "12200\n",
      "12300\n",
      "12400\n",
      "12500\n",
      "12600\n",
      "12700\n",
      "12800\n",
      "12900\n",
      "13000\n",
      "13100\n",
      "13200\n",
      "13300\n",
      "13400\n",
      "13500\n",
      "13600\n",
      "13700\n"
     ]
    }
   ],
   "source": [
    "move_files(abs_dirname, train_dir, test_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "6d53a5c4",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "train_df = pd.DataFrame(columns = [\"path\", \"label\"])\n",
    "test_df = pd.DataFrame(columns = [\"path\", \"label\"])\n",
    "\n",
    "abs_dirname  = \"/Users/harsh/ML/Project/project/dataset/train/parasite\"\n",
    "files = [os.path.join(abs_dirname, f) for f in os.listdir(abs_dirname)]\n",
    "for f in files:\n",
    "    f_base = os.path.basename(f)\n",
    "    if f_base.endswith('.png'):\n",
    "        train_df = train_df.append({'path': f_base, \"label\": 1}, ignore_index=True)\n",
    "\n",
    "abs_dirname  = \"/Users/harsh/ML/Project/project/dataset/train/uninfected\"\n",
    "files = [os.path.join(abs_dirname, f) for f in os.listdir(abs_dirname)]\n",
    "for f in files:\n",
    "    f_base = os.path.basename(f)\n",
    "    if f_base.endswith('.png'):\n",
    "        train_df = train_df.append({'path': f_base, \"label\": 0}, ignore_index=True)\n",
    "    \n",
    "train_df.to_csv(\"/Users/harsh/ML/Project/project/dataset/train.csv\", index = False)\n",
    "\n",
    "\n",
    "abs_dirname  = \"/Users/harsh/ML/Project/project/dataset/test/parasite\"\n",
    "files = [os.path.join(abs_dirname, f) for f in os.listdir(abs_dirname)]\n",
    "for f in files:\n",
    "    f_base = os.path.basename(f)\n",
    "    if f_base.endswith('.png'):\n",
    "        test_df = test_df.append({'path': f_base, \"label\": 1}, ignore_index=True)\n",
    "    \n",
    "abs_dirname  = \"/Users/harsh/ML/Project/project/dataset/test/uninfected\"\n",
    "files = [os.path.join(abs_dirname, f) for f in os.listdir(abs_dirname)]\n",
    "for f in files:\n",
    "    f_base = os.path.basename(f)\n",
    "    if f_base.endswith('.png'):\n",
    "        test_df = test_df.append({'path': f_base, \"label\": 0}, ignore_index=True)\n",
    "    \n",
    "test_df.to_csv(\"/Users/harsh/ML/Project/project/dataset/test.csv\", index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "7e1962d8",
   "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>path</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>C91P52ThinF_IMG_20150821_123116_cell_189.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>C84P45ThinF_IMG_20150818_101226_cell_98.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>C85P46ThinF_IMG_20150820_105154_cell_221.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>C99P60ThinF_IMG_20150918_141620_cell_56.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>C99P60ThinF_IMG_20150918_141620_cell_42.png</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5505</th>\n",
       "      <td>C67P28N_ThinF_IMG_20150819_133000_cell_119.png</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5506</th>\n",
       "      <td>C84P45ThinF_IMG_20150818_102325_cell_1.png</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5507</th>\n",
       "      <td>C95P56ThinF_IMG_20150821_172222_cell_78.png</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5508</th>\n",
       "      <td>C62P23N_ThinF_IMG_20150818_132811_cell_150.png</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5509</th>\n",
       "      <td>C76P37ThinF_IMG_20150815_172808_cell_15.png</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5510 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                path label\n",
       "0       C91P52ThinF_IMG_20150821_123116_cell_189.png     1\n",
       "1        C84P45ThinF_IMG_20150818_101226_cell_98.png     1\n",
       "2       C85P46ThinF_IMG_20150820_105154_cell_221.png     1\n",
       "3        C99P60ThinF_IMG_20150918_141620_cell_56.png     1\n",
       "4        C99P60ThinF_IMG_20150918_141620_cell_42.png     1\n",
       "...                                              ...   ...\n",
       "5505  C67P28N_ThinF_IMG_20150819_133000_cell_119.png     0\n",
       "5506      C84P45ThinF_IMG_20150818_102325_cell_1.png     0\n",
       "5507     C95P56ThinF_IMG_20150821_172222_cell_78.png     0\n",
       "5508  C62P23N_ThinF_IMG_20150818_132811_cell_150.png     0\n",
       "5509     C76P37ThinF_IMG_20150815_172808_cell_15.png     0\n",
       "\n",
       "[5510 rows x 2 columns]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "a35cde03",
   "metadata": {},
   "outputs": [],
   "source": [
    "class CustomImageDataset(Dataset):\n",
    "    def __init__(self, annotations_file, img_dir, transform=None, target_transform=None):\n",
    "        self.img_labels = pd.read_csv(annotations_file)\n",
    "        self.img_dir = img_dir\n",
    "        self.transform = transform\n",
    "        self.target_transform = target_transform\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.img_labels)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        label = self.img_labels['label'][idx]\n",
    "        class_type = \"parasite\" if label == 1 else \"uninfected\"\n",
    "        img_path = os.path.join(self.img_dir, class_type, self.img_labels.iloc[idx, 0])\n",
    "#         print(img_path)\n",
    "        image = cv2.imread(img_path)\n",
    "        image = cv2.resize(image, (32, 32))\n",
    "        image = np.moveaxis(image, -1, 0)\n",
    "        if self.transform:\n",
    "            image = self.transform(image)\n",
    "        if self.target_transform:\n",
    "            label = self.target_transform(label)\n",
    "        return image.astype(np.float32)/255.0, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "8dce54b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dir = \"/Users/harsh/ML/Project/project/dataset/train\"\n",
    "test_dir = \"/Users/harsh/ML/Project/project/dataset/test\"\n",
    "train_csv_path = \"/Users/harsh/ML/Project/project/dataset/train.csv\"\n",
    "test_csv_path = \"/Users/harsh/ML/Project/project/dataset/test.csv\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "bae8ad88",
   "metadata": {},
   "outputs": [],
   "source": [
    "training_data = CustomImageDataset(annotations_file=train_csv_path, img_dir= train_dir)\n",
    "train_dataloader = DataLoader(training_data, batch_size=512, shuffle=True)\n",
    "testing_data = CustomImageDataset(annotations_file=test_csv_path, img_dir= test_dir)\n",
    "test_dataloader = DataLoader(testing_data, batch_size=5510, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a73ed726",
   "metadata": {},
   "source": [
    "<h1>ANN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "af24da3e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Net(\n",
      "  (fc1): Linear(in_features=3072, out_features=256, bias=True)\n",
      "  (fc2): Linear(in_features=256, out_features=128, bias=True)\n",
      "  (fc3): Linear(in_features=128, out_features=64, bias=True)\n",
      "  (fc4): Linear(in_features=64, out_features=1, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "class Net(nn.Module):\n",
    "\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.fc1 = nn.Linear(3072, 256)  \n",
    "        self.fc2 = nn.Linear(256, 128)\n",
    "        self.fc3 = nn.Linear(128, 64)\n",
    "        self.fc4 = nn.Linear(64, 1)\n",
    "\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = torch.flatten(x, start_dim=1)        \n",
    "        x = F.relu(self.fc1(x))\n",
    "        x = F.relu(self.fc2(x))\n",
    "        x = F.relu(self.fc3(x))\n",
    "        x = torch.sigmoid(self.fc4(x))\n",
    "        return x\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "net = Net()\n",
    "print(net)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "c68ce903",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Net(\n",
       "  (fc1): Linear(in_features=3072, out_features=256, bias=True)\n",
       "  (fc2): Linear(in_features=256, out_features=128, bias=True)\n",
       "  (fc3): Linear(in_features=128, out_features=64, bias=True)\n",
       "  (fc4): Linear(in_features=64, out_features=1, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "opt = optim.Adam(net.parameters(), lr = 0.01)\n",
    "loss = []\n",
    "acc_history = []\n",
    "net.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "e0de1a31",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train():\n",
    "     # set model in training mode (need this because of dropout)\n",
    "    loss_history = []\n",
    "    # dataset API gives us pythonic batching \n",
    "    for batch_id, (data, label) in enumerate(train_dataloader):\n",
    "        data = Variable(data)\n",
    "        target = Variable(label)\n",
    "        target = target.unsqueeze(1)\n",
    "        target = target.float()\n",
    "        \n",
    "        # forward pass, calculate loss and backprop!\n",
    "        opt.zero_grad()\n",
    "        preds = net(data)\n",
    "        loss = nn.BCELoss()(preds, target)\n",
    "        loss.backward()\n",
    "        loss_history.append(loss.data.item())\n",
    "        opt.step()\n",
    "        \n",
    "    print(\"Loss: \", sum(loss_history)/len(train_dataloader))\n",
    "    return sum(loss_history)/len(train_dataloader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "8b2ae316",
   "metadata": {},
   "outputs": [],
   "source": [
    "def test(dataloader):\n",
    "    net.eval()\n",
    "    \n",
    "    for data, target in dataloader:\n",
    "        data = Variable(data) \n",
    "        output = net(data)\n",
    "        probability = output.detach().numpy()\n",
    "        probability =  np.where(probability >= 0.5, 1, 0)\n",
    "        target = target.detach().numpy()\n",
    "    return probability, target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "c2440566",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss:  0.6030540859157388\n",
      "Done for iteration:  0\n",
      "Loss:  0.571281836791472\n",
      "Done for iteration:  1\n",
      "Loss:  0.5515619780529629\n",
      "Done for iteration:  2\n",
      "Loss:  0.5500559522347017\n",
      "Done for iteration:  3\n",
      "Loss:  0.5419264232570474\n",
      "Done for iteration:  4\n",
      "Loss:  0.5366736914623867\n",
      "Done for iteration:  5\n",
      "Loss:  0.5448275770653378\n",
      "Done for iteration:  6\n",
      "Loss:  0.555187347937714\n",
      "Done for iteration:  7\n",
      "Loss:  0.5329272401603785\n",
      "Done for iteration:  8\n",
      "Loss:  0.5289103131402623\n",
      "Done for iteration:  9\n",
      "Loss:  0.5522695359858599\n",
      "Done for iteration:  10\n",
      "Loss:  0.5344955020330169\n",
      "Done for iteration:  11\n",
      "Loss:  0.531617864289067\n",
      "Done for iteration:  12\n",
      "Loss:  0.5392983657392588\n",
      "Done for iteration:  13\n",
      "Loss:  0.5218183547258377\n",
      "Done for iteration:  14\n",
      "Loss:  0.5157654400576245\n",
      "Done for iteration:  15\n",
      "Loss:  0.5405205826867711\n",
      "Done for iteration:  16\n",
      "Loss:  0.521846090528098\n",
      "Done for iteration:  17\n",
      "Loss:  0.5194754072211005\n",
      "Done for iteration:  18\n",
      "Loss:  0.5163801888173277\n",
      "Done for iteration:  19\n",
      "Loss:  0.5220332938161764\n",
      "Done for iteration:  20\n",
      "Loss:  0.5190700644796545\n",
      "Done for iteration:  21\n",
      "Loss:  0.5298066681081598\n",
      "Done for iteration:  22\n",
      "Loss:  0.5220586942000822\n",
      "Done for iteration:  23\n",
      "Loss:  0.5177785605192184\n",
      "Done for iteration:  24\n",
      "Loss:  0.5176088938658888\n",
      "Done for iteration:  25\n",
      "Loss:  0.5126945538954302\n",
      "Done for iteration:  26\n",
      "Loss:  0.5255493596196175\n",
      "Done for iteration:  27\n",
      "Loss:  0.5175715806809339\n",
      "Done for iteration:  28\n",
      "Loss:  0.5127165920355103\n",
      "Done for iteration:  29\n",
      "Loss:  0.5278277640992944\n",
      "Done for iteration:  30\n",
      "Loss:  0.5220128643241796\n",
      "Done for iteration:  31\n",
      "Loss:  0.5158630643378604\n",
      "Done for iteration:  32\n",
      "Loss:  0.5126002180305395\n",
      "Done for iteration:  33\n",
      "Loss:  0.5051434297453273\n",
      "Done for iteration:  34\n",
      "Loss:  0.5158604891462759\n",
      "Done for iteration:  35\n",
      "Loss:  0.5182721012017943\n",
      "Done for iteration:  36\n",
      "Loss:  0.5074784268032421\n",
      "Done for iteration:  37\n",
      "Loss:  0.518230174752799\n",
      "Done for iteration:  38\n",
      "Loss:  0.5157058963721449\n",
      "Done for iteration:  39\n",
      "Loss:  0.5129629420963201\n",
      "Done for iteration:  40\n",
      "Loss:  0.5065622383897955\n",
      "Done for iteration:  41\n",
      "Loss:  0.5016960982571949\n",
      "Done for iteration:  42\n",
      "Loss:  0.5109332000667398\n",
      "Done for iteration:  43\n",
      "Loss:  0.4901213693347844\n",
      "Done for iteration:  44\n",
      "Loss:  0.5026927786794576\n",
      "Done for iteration:  45\n",
      "Loss:  0.4926116940650073\n",
      "Done for iteration:  46\n",
      "Loss:  0.5088475488803603\n",
      "Done for iteration:  47\n",
      "Loss:  0.4980589131062681\n",
      "Done for iteration:  48\n",
      "Loss:  0.5077070180665363\n",
      "Done for iteration:  49\n",
      "Loss:  0.5019071007316763\n",
      "Done for iteration:  50\n",
      "Loss:  0.5113543888384645\n",
      "Done for iteration:  51\n",
      "Loss:  0.5056750002232465\n",
      "Done for iteration:  52\n",
      "Loss:  0.5002164332704111\n",
      "Done for iteration:  53\n",
      "Loss:  0.5196695476770401\n",
      "Done for iteration:  54\n",
      "Loss:  0.49083931811831216\n",
      "Done for iteration:  55\n",
      "Loss:  0.49278479340401565\n",
      "Done for iteration:  56\n",
      "Loss:  0.5027002157135443\n",
      "Done for iteration:  57\n",
      "Loss:  0.5224594839594581\n",
      "Done for iteration:  58\n",
      "Loss:  0.4850788820873607\n",
      "Done for iteration:  59\n",
      "Loss:  0.49020869691263547\n",
      "Done for iteration:  60\n",
      "Loss:  0.4881048046729781\n",
      "Done for iteration:  61\n",
      "Loss:  0.5252878178249706\n",
      "Done for iteration:  62\n",
      "Loss:  0.5060338147661902\n",
      "Done for iteration:  63\n",
      "Loss:  0.5092214941978455\n",
      "Done for iteration:  64\n",
      "Loss:  0.5026680332693186\n",
      "Done for iteration:  65\n",
      "Loss:  0.50659609653733\n",
      "Done for iteration:  66\n",
      "Loss:  0.5198734646493738\n",
      "Done for iteration:  67\n",
      "Loss:  0.5024770572781563\n",
      "Done for iteration:  68\n",
      "Loss:  0.5035185827450319\n",
      "Done for iteration:  69\n",
      "Loss:  0.49945490197701886\n",
      "Done for iteration:  70\n",
      "Loss:  0.49666898900812323\n",
      "Done for iteration:  71\n",
      "Loss:  0.48786932500925934\n",
      "Done for iteration:  72\n",
      "Loss:  0.48931943557479163\n",
      "Done for iteration:  73\n",
      "Loss:  0.5116614685817198\n",
      "Done for iteration:  74\n",
      "Loss:  0.495826816694303\n",
      "Done for iteration:  75\n",
      "Loss:  0.4934321655468507\n",
      "Done for iteration:  76\n",
      "Loss:  0.4849553873593157\n",
      "Done for iteration:  77\n",
      "Loss:  0.4808864336122166\n",
      "Done for iteration:  78\n",
      "Loss:  0.47995618053457956\n",
      "Done for iteration:  79\n",
      "Loss:  0.48153336481614545\n",
      "Done for iteration:  80\n",
      "Loss:  0.48154179548675363\n",
      "Done for iteration:  81\n",
      "Loss:  0.48984390971335495\n",
      "Done for iteration:  82\n",
      "Loss:  0.4838955497199839\n",
      "Done for iteration:  83\n",
      "Loss:  0.47827721725810657\n",
      "Done for iteration:  84\n",
      "Loss:  0.4878874353387139\n",
      "Done for iteration:  85\n",
      "Loss:  0.48136222633448517\n",
      "Done for iteration:  86\n",
      "Loss:  0.5063026052984324\n",
      "Done for iteration:  87\n",
      "Loss:  0.4996234666217457\n",
      "Done for iteration:  88\n",
      "Loss:  0.5108603245832704\n",
      "Done for iteration:  89\n",
      "Loss:  0.48600214652039786\n",
      "Done for iteration:  90\n",
      "Loss:  0.4883426976474849\n",
      "Done for iteration:  91\n",
      "Loss:  0.4818204553289847\n",
      "Done for iteration:  92\n",
      "Loss:  0.49337751282887027\n",
      "Done for iteration:  93\n",
      "Loss:  0.49640367992899637\n",
      "Done for iteration:  94\n",
      "Loss:  0.48325996642762964\n",
      "Done for iteration:  95\n",
      "Loss:  0.48462970554828644\n",
      "Done for iteration:  96\n",
      "Loss:  0.4812599908221852\n",
      "Done for iteration:  97\n",
      "Loss:  0.4829485233534466\n",
      "Done for iteration:  98\n",
      "Loss:  0.4814532683654265\n",
      "Done for iteration:  99\n"
     ]
    }
   ],
   "source": [
    "for i in range(0, 100):\n",
    "    loss.append(train())\n",
    "    print(\"Done for iteration: \", i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "2e29e38d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "100"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss_100 = copy.deepcopy(loss)\n",
    "len(loss_100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "6ed44a66",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss:  0.519852876663208\n",
      "Done for iteration:  0\n",
      "Loss:  0.5318189263343811\n",
      "Done for iteration:  1\n",
      "Loss:  0.5122197866439819\n",
      "Done for iteration:  2\n",
      "Loss:  0.4845896065235138\n",
      "Done for iteration:  3\n",
      "Loss:  0.4725850820541382\n",
      "Done for iteration:  4\n",
      "Loss:  0.4828445017337799\n",
      "Done for iteration:  5\n",
      "Loss:  0.49535778164863586\n",
      "Done for iteration:  6\n",
      "Loss:  0.49663159251213074\n",
      "Done for iteration:  7\n",
      "Loss:  0.4850935637950897\n",
      "Done for iteration:  8\n",
      "Loss:  0.47800979018211365\n",
      "Done for iteration:  9\n",
      "Loss:  0.47809988260269165\n",
      "Done for iteration:  10\n",
      "Loss:  0.4745863080024719\n",
      "Done for iteration:  11\n",
      "Loss:  0.475283682346344\n",
      "Done for iteration:  12\n",
      "Loss:  0.4767259955406189\n",
      "Done for iteration:  13\n",
      "Loss:  0.4759446680545807\n",
      "Done for iteration:  14\n",
      "Loss:  0.47717559337615967\n",
      "Done for iteration:  15\n",
      "Loss:  0.47181031107902527\n",
      "Done for iteration:  16\n",
      "Loss:  0.46242213249206543\n",
      "Done for iteration:  17\n",
      "Loss:  0.46080151200294495\n",
      "Done for iteration:  18\n",
      "Loss:  0.4624778628349304\n",
      "Done for iteration:  19\n",
      "Loss:  0.46646848320961\n",
      "Done for iteration:  20\n",
      "Loss:  0.4640652537345886\n",
      "Done for iteration:  21\n",
      "Loss:  0.4589405655860901\n",
      "Done for iteration:  22\n",
      "Loss:  0.4587695598602295\n",
      "Done for iteration:  23\n",
      "Loss:  0.4584614932537079\n",
      "Done for iteration:  24\n",
      "Loss:  0.4588333070278168\n",
      "Done for iteration:  25\n",
      "Loss:  0.45746394991874695\n",
      "Done for iteration:  26\n",
      "Loss:  0.4551483392715454\n",
      "Done for iteration:  27\n",
      "Loss:  0.45621272921562195\n",
      "Done for iteration:  28\n",
      "Loss:  0.45528656244277954\n",
      "Done for iteration:  29\n",
      "Loss:  0.45644715428352356\n",
      "Done for iteration:  30\n",
      "Loss:  0.4537811577320099\n",
      "Done for iteration:  31\n",
      "Loss:  0.45366430282592773\n",
      "Done for iteration:  32\n",
      "Loss:  0.45270711183547974\n",
      "Done for iteration:  33\n",
      "Loss:  0.45282313227653503\n",
      "Done for iteration:  34\n",
      "Loss:  0.4532352089881897\n",
      "Done for iteration:  35\n",
      "Loss:  0.4517776072025299\n",
      "Done for iteration:  36\n",
      "Loss:  0.4512164890766144\n",
      "Done for iteration:  37\n",
      "Loss:  0.4505879282951355\n",
      "Done for iteration:  38\n",
      "Loss:  0.45118582248687744\n",
      "Done for iteration:  39\n",
      "Loss:  0.4505077600479126\n",
      "Done for iteration:  40\n",
      "Loss:  0.44982436299324036\n",
      "Done for iteration:  41\n",
      "Loss:  0.4494219720363617\n",
      "Done for iteration:  42\n",
      "Loss:  0.4495607018470764\n",
      "Done for iteration:  43\n",
      "Loss:  0.44981396198272705\n",
      "Done for iteration:  44\n",
      "Loss:  0.44919559359550476\n",
      "Done for iteration:  45\n",
      "Loss:  0.4486888349056244\n",
      "Done for iteration:  46\n",
      "Loss:  0.44832682609558105\n",
      "Done for iteration:  47\n",
      "Loss:  0.4485575556755066\n",
      "Done for iteration:  48\n",
      "Loss:  0.44840478897094727\n",
      "Done for iteration:  49\n",
      "Loss:  0.448151171207428\n",
      "Done for iteration:  50\n",
      "Loss:  0.44778168201446533\n",
      "Done for iteration:  51\n",
      "Loss:  0.4476514458656311\n",
      "Done for iteration:  52\n",
      "Loss:  0.4476720988750458\n",
      "Done for iteration:  53\n",
      "Loss:  0.44742584228515625\n",
      "Done for iteration:  54\n",
      "Loss:  0.44713160395622253\n",
      "Done for iteration:  55\n",
      "Loss:  0.44691815972328186\n",
      "Done for iteration:  56\n",
      "Loss:  0.44693928956985474\n",
      "Done for iteration:  57\n",
      "Loss:  0.44672951102256775\n",
      "Done for iteration:  58\n",
      "Loss:  0.4465561509132385\n",
      "Done for iteration:  59\n",
      "Loss:  0.4463970363140106\n",
      "Done for iteration:  60\n",
      "Loss:  0.4463558495044708\n",
      "Done for iteration:  61\n",
      "Loss:  0.446133017539978\n",
      "Done for iteration:  62\n",
      "Loss:  0.4459618330001831\n",
      "Done for iteration:  63\n",
      "Loss:  0.4458473324775696\n",
      "Done for iteration:  64\n",
      "Loss:  0.445766806602478\n",
      "Done for iteration:  65\n",
      "Loss:  0.44562652707099915\n",
      "Done for iteration:  66\n",
      "Loss:  0.44551798701286316\n",
      "Done for iteration:  67\n",
      "Loss:  0.44536489248275757\n",
      "Done for iteration:  68\n",
      "Loss:  0.4452475607395172\n",
      "Done for iteration:  69\n",
      "Loss:  0.4450899362564087\n",
      "Done for iteration:  70\n",
      "Loss:  0.44502177834510803\n",
      "Done for iteration:  71\n",
      "Loss:  0.4448848068714142\n",
      "Done for iteration:  72\n",
      "Loss:  0.4447602927684784\n",
      "Done for iteration:  73\n",
      "Loss:  0.44463565945625305\n",
      "Done for iteration:  74\n",
      "Loss:  0.4445345103740692\n",
      "Done for iteration:  75\n",
      "Loss:  0.44442760944366455\n",
      "Done for iteration:  76\n",
      "Loss:  0.44431042671203613\n",
      "Done for iteration:  77\n",
      "Loss:  0.4442307949066162\n",
      "Done for iteration:  78\n",
      "Loss:  0.4441128671169281\n",
      "Done for iteration:  79\n",
      "Loss:  0.444008469581604\n",
      "Done for iteration:  80\n",
      "Loss:  0.44390931725502014\n",
      "Done for iteration:  81\n",
      "Loss:  0.4438225030899048\n",
      "Done for iteration:  82\n",
      "Loss:  0.443704754114151\n",
      "Done for iteration:  83\n",
      "Loss:  0.4436131417751312\n",
      "Done for iteration:  84\n",
      "Loss:  0.4435144066810608\n",
      "Done for iteration:  85\n",
      "Loss:  0.4434095025062561\n",
      "Done for iteration:  86\n",
      "Loss:  0.4433177709579468\n",
      "Done for iteration:  87\n",
      "Loss:  0.44321733713150024\n",
      "Done for iteration:  88\n",
      "Loss:  0.4431188702583313\n",
      "Done for iteration:  89\n",
      "Loss:  0.44303053617477417\n",
      "Done for iteration:  90\n",
      "Loss:  0.4429295063018799\n",
      "Done for iteration:  91\n",
      "Loss:  0.44283509254455566\n",
      "Done for iteration:  92\n",
      "Loss:  0.4427485466003418\n",
      "Done for iteration:  93\n",
      "Loss:  0.4426548480987549\n",
      "Done for iteration:  94\n",
      "Loss:  0.4425621032714844\n",
      "Done for iteration:  95\n",
      "Loss:  0.44247424602508545\n",
      "Done for iteration:  96\n",
      "Loss:  0.44238489866256714\n",
      "Done for iteration:  97\n",
      "Loss:  0.4422950744628906\n",
      "Done for iteration:  98\n",
      "Loss:  0.44220802187919617\n",
      "Done for iteration:  99\n"
     ]
    }
   ],
   "source": [
    "loss = []\n",
    "for i in range(0, 100):\n",
    "    loss.append(train())\n",
    "    print(\"Done for iteration: \", i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "c82296b1",
   "metadata": {},
   "outputs": [],
   "source": [
    "accuracy_df = pd.DataFrame(columns = ['Accuracy', 'Precision', 'Recall','F1'], \n",
    "                   index = ['Train', 'Test'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "b960ef81",
   "metadata": {},
   "outputs": [],
   "source": [
    "training_data = CustomImageDataset(annotations_file=train_csv_path, img_dir= train_dir)\n",
    "train_dataloader = DataLoader(training_data, batch_size=22048, shuffle=True)\n",
    "testing_data = CustomImageDataset(annotations_file=test_csv_path, img_dir= test_dir)\n",
    "test_dataloader = DataLoader(testing_data, batch_size=8267, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "0a0bffd3",
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred, y_test = test(test_dataloader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "969246e1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1483, 1272],\n",
       "       [ 730, 2025]])"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "matrix,accuracy,precision,recall,f1 = metric(y_test, y_pred)\n",
    "accuracy_df.loc['Train'] = [accuracy,precision,recall,f1]\n",
    "matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "0f315240",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1483, 1272],\n",
       "       [ 730, 2025]])"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred, y_test = test(train_dataloader)\n",
    "accuracy_df.loc['Test'] = [accuracy,precision,recall,f1]\n",
    "matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "c774817d",
   "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>Accuracy</th>\n",
       "      <th>Precision</th>\n",
       "      <th>Recall</th>\n",
       "      <th>F1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Train</th>\n",
       "      <td>0.636661</td>\n",
       "      <td>0.614195</td>\n",
       "      <td>0.735027</td>\n",
       "      <td>0.6692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Test</th>\n",
       "      <td>0.636661</td>\n",
       "      <td>0.614195</td>\n",
       "      <td>0.735027</td>\n",
       "      <td>0.6692</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       Accuracy Precision    Recall      F1\n",
       "Train  0.636661  0.614195  0.735027  0.6692\n",
       "Test   0.636661  0.614195  0.735027  0.6692"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "b8a4c090",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fbbf86417c0>]"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.plot(list(range(1, 201)), loss_200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7d4aec16",
   "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": "3.8.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}