detection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "e1f036a1",
   "metadata": {},
   "source": [
    "<strong>Step1: Brief description of the problem and data (5 pts)</strong></br>\n",
    "<strong>Briefly describe the challenge problem and NLP. Describe the size, dimension, structure, etc., of the data. </strong>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "684a4d30",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "e4d64d9a",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "220025\n",
      "57458\n"
     ]
    }
   ],
   "source": [
    "print(len(os.listdir('/hpctmp/e0427773/histopathologic-cancer-detection/train/')))\n",
    "print(len(os.listdir('/hpctmp/e0427773/histopathologic-cancer-detection/test/')))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d905fa72",
   "metadata": {},
   "source": [
    "<strong>There are 220,025 training samples and 57,458 test samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "203ea30f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(220025, 2)\n"
     ]
    }
   ],
   "source": [
    "train_labels = pd.read_csv('/hpctmp/e0427773/histopathologic-cancer-detection/train_labels.csv', dtype=str)\n",
    "print(train_labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "16cee3d4",
   "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>id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>f38a6374c348f90b587e046aac6079959adf3835</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>c18f2d887b7ae4f6742ee445113fa1aef383ed77</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>755db6279dae599ebb4d39a9123cce439965282d</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>bc3f0c64fb968ff4a8bd33af6971ecae77c75e08</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>068aba587a4950175d04c680d38943fd488d6a9d</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                         id  label\n",
       "0  f38a6374c348f90b587e046aac6079959adf3835    0.0\n",
       "1  c18f2d887b7ae4f6742ee445113fa1aef383ed77    1.0\n",
       "2  755db6279dae599ebb4d39a9123cce439965282d    0.0\n",
       "3  bc3f0c64fb968ff4a8bd33af6971ecae77c75e08    0.0\n",
       "4  068aba587a4950175d04c680d38943fd488d6a9d    0.0"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels['label'] = train_labels['label'].astype(float)\n",
    "train_labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c0f33c47",
   "metadata": {},
   "source": [
    "<strong>Step2: Exploratory Data Analysis (EDA) — Inspect, Visualize and Clean the Data (15 pts)</strong></br>\n",
    "<strong>Show a few visualizations like histograms. Describe any data cleaning procedures. Based on your EDA, what is your plan of analysis?</strong>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "841f6f24",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0    130908\n",
       "1.0     89117\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels['label'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "cf69e6be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:ylabel='label'>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_labels['label'].value_counts().plot(kind='pie')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e841b5b0",
   "metadata": {},
   "source": [
    "<strong>There are 130,908 non-cancer images and 89,117 cancer images in the training samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2ea1b997",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels_pos = train_labels[train_labels['label']==1]\n",
    "train_labels_neg = train_labels[train_labels['label']==0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c9188800",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_labels_neg = train_labels_neg.sample(n = train_labels_pos.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "074d5524",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(178234, 2)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels_equal = pd.concat([train_labels_neg,train_labels_pos]).sample(frac=1, random_state=17).reset_index(drop=True)\n",
    "train_labels_equal.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "5b9adc53",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:ylabel='label'>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_labels_equal['label'].value_counts().plot(kind='pie')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "579f1fbf",
   "metadata": {},
   "source": [
    "<strong>A balanced dataset was created by combining equal numbers of cancer (pos) and non-cancer (neg) images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "838b5132",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "2b25a86f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from PIL import Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "74abaaaa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<PIL.TiffImagePlugin.TiffImageFile image mode=RGB size=96x96 at 0x2B77CDC887C0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(96, 96)\n"
     ]
    }
   ],
   "source": [
    "with Image.open('/hpctmp/e0427773/histopathologic-cancer-detection/train/7ff2b0252621ce92abe5cad8afdf331c7000612a.tif') as im:\n",
    "    im.show()\n",
    "    print(im.size)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d51df688",
   "metadata": {},
   "source": [
    "<strong>A sample image is shown here and the size is 96 pixels in height and 96 pixels in width"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3fea27a5",
   "metadata": {},
   "source": [
    "<strong>Step3: Model Architecture (25 pts)</br>\n",
    "Describe your model architecture and reasoning for why you believe that specific architecture would be suitable for this problem. Compare multiple architectures and tune hyperparameters. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "f7c41b66",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "95aa2073",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_img, valid_img = train_test_split(train_labels_equal, test_size=0.25,\n",
    "                                        random_state=1710, stratify=train_labels_equal.label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "6d4f22d5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow import keras\n",
    "from keras.layers import Dense, Activation, Flatten, Dropout, BatchNormalization\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras import regularizers, optimizers\n",
    "from keras.layers import PReLU\n",
    "from keras.initializers import Constant\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "8bf9a4ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_img['id'] = train_img['id']+'.tif'\n",
    "valid_img['id'] = valid_img['id']+'.tif'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "eb3b3e40",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_img['label'] = train_img['label'].astype(str)\n",
    "valid_img['label'] = valid_img['label'].astype(str)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2a630e9c",
   "metadata": {},
   "source": [
    "<strong>The codes below were uploaded and run on HPC due to the limited computational resource of the local machine"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6783326b",
   "metadata": {},
   "source": [
    "<strong>First I'd like to try a model with four hidden layers. Each hidden layer has two convolutional neural networks and one max pooling layer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "10972c10",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_datagen=ImageDataGenerator(rescale=1/255)\n",
    "\n",
    "train_gen=train_datagen.flow_from_dataframe(dataframe=train_img,directory=\"/hpctmp/e0427773/histopathologic-cancer-detection/train/\",\n",
    "                x_col=\"id\",y_col=\"label\",batch_size=64,seed=1710,shuffle=True,\n",
    "                class_mode=\"binary\",target_size=(96,96))\n",
    "\n",
    "valid_gen=train_datagen.flow_from_dataframe(dataframe=valid_img,directory=\"/hpctmp/e0427773/histopathologic-cancer-detection/train/\",\n",
    "                x_col=\"id\",y_col=\"label\",batch_size=64,seed=1710,shuffle=True,\n",
    "                class_mode=\"binary\",target_size=(96,96))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9375fa98",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Conv2D(32, (3, 3), padding='same',input_shape=(96,96,3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model.add(Conv2D(128, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(128, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "model.add(Conv2D(256, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Conv2D(256, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model.add(BatchNormalization())\n",
    "          \n",
    "model.add(Flatten())\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(512))\n",
    "model.add(Activation('relu'))\n",
    "\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(256))\n",
    "model.add(Activation('relu'))\n",
    "\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu')) \n",
    "\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "opt = tf.keras.optimizers.Adam(0.001)\n",
    "model.compile(loss='binary_crossentropy', optimizer=opt, metrics=['accuracy'])\n",
    "\n",
    "hist_df = pd.DataFrame(history.history) \n",
    "with open('history.csv', mode='w') as f:\n",
    "    hist_df.to_csv(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "404f4181",
   "metadata": {},
   "source": [
    "<strong>Then I'd like to increase the complexity of each hidden layer but reduce the total number of hidden layers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4ef8c1ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "model2 = Sequential()\n",
    "model2.add(Conv2D(32, (3, 3), padding='same',input_shape=(96,96,3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(32, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(32, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(32, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model2.add(BatchNormalization())\n",
    "\n",
    "model2.add(Conv2D(64, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(64, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(64, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(64, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model2.add(BatchNormalization())\n",
    "\n",
    "model2.add(Conv2D(128, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(128, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(128, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(Conv2D(128, (3, 3)))\n",
    "model2.add(Activation('relu'))\n",
    "model2.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "model2.add(BatchNormalization())\n",
    "          \n",
    "model2.add(Flatten())\n",
    "model2.add(Dropout(0.25))\n",
    "model2.add(Dense(256))\n",
    "model2.add(Activation('relu'))\n",
    "\n",
    "model2.add(Dropout(0.25))\n",
    "model2.add(Dense(128))\n",
    "model2.add(Activation('relu'))\n",
    "\n",
    "model2.add(Dropout(0.25))\n",
    "model2.add(Dense(64))\n",
    "model2.add(Activation('relu')) \n",
    "\n",
    "model2.add(Dropout(0.25))\n",
    "model2.add(Dense(1, activation='sigmoid'))\n",
    "opt = tf.keras.optimizers.Adam(0.001)\n",
    "model2.compile(loss='binary_crossentropy', optimizer=opt, metrics=['accuracy'])\n",
    "\n",
    "model2.summary()\n",
    "\n",
    "STEP_SIZE_TRAIN=train_generator.n//train_generator.batch_size\n",
    "STEP_SIZE_VALID=valid_generator.n//valid_generator.batch_size\n",
    "\n",
    "history2 = model2.fit_generator(generator=train_generator,\n",
    "                    steps_per_epoch=STEP_SIZE_TRAIN,\n",
    "                    validation_data=valid_generator,\n",
    "                    validation_steps=STEP_SIZE_VALID,\n",
    "                    epochs=30, verbose=1\n",
    ")\n",
    "\n",
    "hist2_df = pd.DataFrame(history2.history) \n",
    "with open('history2.csv', mode='w') as f:\n",
    "    hist2_df.to_csv(f)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f8dcec4a",
   "metadata": {},
   "source": [
    "<strong>Step4: Results and Analysis (35 pts)</br>\n",
    "Run hyperparameter tuning, try different architectures for comparison, apply techniques to improve training or performance, and discuss what helped.\n",
    "Includes results with tables and figures. There is an analysis of why or why not something worked well, troubleshooting, and a hyperparameter optimization procedure summary."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "60e22ce1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fa8bb1b2550>]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(hist1.index+1, hist1['accuracy'], color='red')\n",
    "plt.plot(hist1.index+1, hist1['val_accuracy'], color='red', linestyle='dashed')\n",
    "plt.plot(hist2.index+1, hist2['accuracy'], color='blue')\n",
    "plt.plot(hist2.index+1, hist2['val_accuracy'], color='blue', linestyle='dashed')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14bcee24",
   "metadata": {},
   "source": [
    "<strong>Model 1 outperformed model 2 regarding the accuracy"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c61071db",
   "metadata": {},
   "source": [
    "<strong>Strp5: Conclusion (15 pts)</br>\n",
    "Discuss and interpret results as well as learnings and takeaways. What did and did not help improve the performance of your models? What improvements could you try in the future?"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18d170d1",
   "metadata": {},
   "source": [
    "<strong>Various problems require different complexity of the model. Always try different numbers of hidden layers and different numbers of convolutional layers within each hidden layer. Meanwhile, it is important to have a balanced dataset for training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9b7fc868",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set = os.listdir('/hpctmp/e0427773/histopathologic-cancer-detection/test/')\n",
    "test_df = pd.DataFrame(test_set)\n",
    "test_df.columns = ['id']\n",
    "test_datagen=ImageDataGenerator(rescale=1/255)\n",
    "\n",
    "test_generator=test_datagen.flow_from_dataframe(dataframe=test_df, directory=\"/hpctmp/e0427773/histopathologic-cancer-detection/test/\",x_col=\"id\",batch_size=64,seed=1710,shuffle=False,class_mode=None,target_size=(96,96))\n",
    "\n",
    "STEP_SIZE_TEST=test_generator.n//test_generator.batch_size + 1\n",
    "\n",
    "preds = model.predict_generator(generator=test_generator,steps=STEP_SIZE_TEST, verbose = 1)\n",
    "\n",
    "predictions = []\n",
    "for pred in preds:\n",
    "    if pred >= 0.5:\n",
    "        predictions.append(1)\n",
    "    else:\n",
    "        predictions.append(0)\n",
    "\n",
    "submission = test_df.copy()\n",
    "submission['id'] = list(map(lambda x: x[:-4], submission['id']))\n",
    "submission['label']=predictions\n",
    "submission.to_csv('submission.csv',index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.9.4 64-bit",
   "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": "3.9.4"
  },
  "vscode": {
   "interpreter": {
    "hash": "7812ea015bdcee6f23a998adcdd2ef97c151c0c241b7b7070987d9313e41299d"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}