Notebooks/Eye Disorder Classification with Residual Attention Network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "import sys\n",
    "sys.path.append(\"../\")\n",
    "\n",
    "from Code.ResidualAttentionNetwork import ResidualAttentionNetwork\n",
    "import numpy as np\n",
    "import pandas as pd \n",
    "import os\n",
    "import glob\n",
    "from datetime import datetime\n",
    "from pathlib import Path\n",
    "\n",
    "from sklearn.utils import resample\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.utils import class_weight\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint, CSVLogger\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import load_model\n",
    "from tensorflow.keras import optimizers \n",
    "\n",
    "import logging\n",
    "\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'  # FATAL\n",
    "logging.getLogger('tensorflow').setLevel(logging.FATAL)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Organize Data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Construct the files to be placed inside a dataframe\n",
    "def construct_data(filepath):\n",
    "    data = {}\n",
    "    data['filepath'] = filepath\n",
    "    data['class'] = str(Path(filepath).parent).split('/')[-1]\n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dir = \"/pylon5/cc5614p/deopha32/eye_images/\"\n",
    "\n",
    "data_files=[filename for filename in glob.iglob(data_dir + '*/*/*', recursive=True)]\n",
    "\n",
    "# Place inside dataframe\n",
    "data = [construct_data(file) for file in data_files]\n",
    "eye_df = pd.DataFrame(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = eye_df['filepath']\n",
    "y = eye_df['class']\n",
    "\n",
    "# 60-20-20 Split: https://stackoverflow.com/a/38251213/9221241\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, stratify=y)\n",
    "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.25, stratify=y_train)\n",
    "\n",
    "model_train_data = pd.concat([X_train, y_train], axis=1)\n",
    "model_train_data = pd.DataFrame(model_train_data)\n",
    "\n",
    "model_val_data = pd.concat([X_val, y_val], axis=1)\n",
    "model_val_data = pd.DataFrame(model_val_data)\n",
    "\n",
    "model_test_data = pd.concat([X_test, y_test], axis=1)\n",
    "model_test_data = pd.DataFrame(model_test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_names = ['CNV', 'DME', 'DRUSEN', 'NORMAL']\n",
    "\n",
    "weights = class_weight.compute_class_weight('balanced',\n",
    "                                             class_names,\n",
    "                                             y_train)\n",
    "\n",
    "class_weights = {class_names[index]:weights[index] for index in range(len(class_names))}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save Dataframes for future use\n",
    "model_train_data.to_csv(\"/home/deopha32/EyeDiseaseClassification/Data/training_data\", index=False)\n",
    "model_val_data.to_csv(\"/home/deopha32/EyeDiseaseClassification/Data/validatation_data\", index=False)\n",
    "model_test_data.to_csv(\"/home/deopha32/EyeDiseaseClassification/Data/testing_data\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Network Metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGE_WIDTH=32\n",
    "IMAGE_HEIGHT=32\n",
    "IMAGE_SIZE=(IMAGE_WIDTH, IMAGE_HEIGHT)\n",
    "IMAGE_CHANNELS=1\n",
    "IMAGE_SHAPE=(IMAGE_WIDTH, IMAGE_HEIGHT, IMAGE_CHANNELS)\n",
    "\n",
    "batch_size=32\n",
    "\n",
    "num_classes = 4\n",
    "\n",
    "epochs = 500"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Image Generators"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_datagen = ImageDataGenerator(rescale=1./255.)\n",
    "\n",
    "train_generator=train_datagen.flow_from_dataframe(\n",
    "    dataframe=model_train_data,\n",
    "    x_col=\"filepath\",\n",
    "    y_col=\"class\",\n",
    "    batch_size=batch_size,\n",
    "    shuffle=True,\n",
    "    class_mode=\"categorical\",\n",
    "    target_size=IMAGE_SIZE,\n",
    "    color_mode='grayscale',\n",
    "    validate_filenames=False\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_datagen = ImageDataGenerator(rescale=1./255.)\n",
    "\n",
    "valid_generator = valid_datagen.flow_from_dataframe(\n",
    "    dataframe=model_val_data,\n",
    "    x_col=\"filepath\",\n",
    "    y_col=\"class\",\n",
    "    batch_size=batch_size,\n",
    "    shuffle=True,\n",
    "    class_mode=\"categorical\",\n",
    "    target_size=IMAGE_SIZE,\n",
    "    color_mode='grayscale',\n",
    "    validate_filenames=False\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_indices = train_generator.class_indices\n",
    "new_class_weights = dict((class_indices[key], value) for (key, value) in class_weights.items())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "curr_time = f'{datetime.now():%H-%M-%S%z_%m%d%Y}'\n",
    "\n",
    "model_path = \"/pylon5/cc5614p/deopha32/Saved_Models/eye-model_{}.h5\".format(curr_time)\n",
    "logger_path = \"/pylon5/cc5614p/deopha32/Saved_Models/eye-model-history_{}.csv\".format(curr_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "checkpoint = ModelCheckpoint(model_path, monitor='val_acc', verbose=1, save_best_only=True)\n",
    "csv_logger = CSVLogger(logger_path, append=True)\n",
    "\n",
    "callbacks = [checkpoint, csv_logger]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "STEP_SIZE_TRAIN=train_generator.n//train_generator.batch_size\n",
    "STEP_SIZE_VALID=valid_generator.n//valid_generator.batch_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model Training\n",
    "with tf.device('/gpu:0'):\n",
    "    model = ResidualAttentionNetwork(\n",
    "                input_shape=IMAGE_SHAPE, \n",
    "                n_classes=num_classes, \n",
    "                activation='softmax').build_model()\n",
    "    \n",
    "    model.compile(optimizer=optimizers.RMSprop(lr=0.0001),\n",
    "                  loss='categorical_crossentropy',\n",
    "                  metrics=['accuracy'])\n",
    "    \n",
    "    history = model.fit_generator(generator=train_generator, class_weight=new_class_weights,\n",
    "                    steps_per_epoch=STEP_SIZE_TRAIN, verbose=1, callbacks=callbacks,\n",
    "                    validation_data=valid_generator, validation_steps=STEP_SIZE_VALID,\n",
    "                    epochs=epochs, use_multiprocessing=True, workers=40)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}