train.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of Copy of Copy of train.ipynb",
      "private_outputs": true,
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/cbalkig/Anomaly_Detection_in_Videos/blob/master/train.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z5l0wNtKs7vP"
      },
      "source": [
        "import os\n",
        "import numpy as np\n",
        "import keras\n",
        "from keras.layers import Conv2D, Flatten, Dense, Dropout, MaxPooling2D\n",
        "from keras.models import Sequential, load_model\n",
        "import matplotlib.pyplot as plt\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "import time"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hDd3-dzSs82m"
      },
      "source": [
        "working_directory = '/content/drive/MyDrive/AnomalyDetectionInVideos'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pKtCPjOYs97U"
      },
      "source": [
        "class Config:\n",
        "    EDIT_DATASET_PATH = os.path.join(working_directory, \"regenerated_files\")\n",
        "    BATCH_SIZE = 256\n",
        "    EPOCHS = 10\n",
        "    MODEL_PATH = os.path.join(working_directory, \"model.hdf5\")\n",
        "    IMAGE_SIZE = 256"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t9tnVYvkQR6x"
      },
      "source": [
        "def show_batch(image_batch, label_batch):\n",
        "    plt.figure(figsize=(10,10))\n",
        "    for n in range(10):\n",
        "        ax = plt.subplot(5,5,n+1)\n",
        "        plt.imshow(image_batch[n, :, :, 0], cmap='gray')\n",
        "        plt.axis('off')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rxkazTWsRyIb"
      },
      "source": [
        "def plot_acc_loss(trained):\n",
        "    fig, ax = plt.subplots(1, 2, figsize=(15,5))\n",
        "    ax[0].set_title('loss')\n",
        "    ax[0].plot(trained.epoch, trained.history[\"loss\"], label=\"Train loss\")\n",
        "    ax[1].set_title('acc')\n",
        "    ax[1].plot(trained.epoch, trained.history[\"accuracy\"], label=\"Train acc\")\n",
        "    ax[0].legend()\n",
        "    ax[1].legend()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zxThkfDO3jDe"
      },
      "source": [
        "def train_model(reload_model=True):\n",
        "    if reload_model:\n",
        "        return load_model(Config.MODEL_PATH)\n",
        "    input_shape = (Config.IMAGE_SIZE, Config.IMAGE_SIZE, 1)\n",
        "\n",
        "    model = Sequential()\n",
        "\n",
        "    model.add(Conv2D(32, kernel_size=3, activation=\"relu\", input_shape=input_shape))\n",
        "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "    model.add(Conv2D(32, kernel_size=3, activation=\"relu\", input_shape=input_shape))\n",
        "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "    model.add(Conv2D(64, kernel_size=3, activation=\"relu\", input_shape=input_shape))\n",
        "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "    model.add(Flatten())\n",
        "    model.add(Dense(64, activation=\"relu\"))\n",
        "    model.add(Dropout(0.5))\n",
        "    model.add(Dense(1, activation=\"sigmoid\"))\n",
        "\n",
        "    model.summary()\n",
        "\n",
        "    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])\n",
        "    history = model.fit(train_generator, epochs=Config.EPOCHS, verbose = 1)\n",
        "    plot_acc_loss(history)\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gVPZox-2cXMD"
      },
      "source": [
        "start_time = time.time()\n",
        "print(\"Sample Count\", sum(len(files) for _, _, files in os.walk(Config.EDIT_DATASET_PATH)))\n",
        "train_datagen = ImageDataGenerator(\n",
        "        rescale=1./255,\n",
        "        shear_range=0.2,\n",
        "        rotation_range=30, \n",
        "        fill_mode='nearest',\n",
        "        brightness_range=[0.4,1.5])\n",
        "train_generator = train_datagen.flow_from_directory(\n",
        "        Config.EDIT_DATASET_PATH,\n",
        "        target_size=(Config.IMAGE_SIZE, Config.IMAGE_SIZE),\n",
        "        batch_size=Config.BATCH_SIZE,\n",
        "        shuffle=True,\n",
        "        color_mode = 'grayscale',\n",
        "        class_mode='binary')\n",
        "print(\"Sample Count after ImageDataGenerator\", train_generator.samples)\n",
        "\n",
        "image_batch, label_batch = next(train_generator)\n",
        "show_batch(image_batch, label_batch)\n",
        "print(\"ImageDataGenerator Execution time:\", time.time() - start_time, \"seconds.\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Fe_09tfotJjQ"
      },
      "source": [
        "start_time = time.time()\n",
        "model = train_model(reload_model=False)\n",
        "print(\"Train Execution time:\", time.time() - start_time, \"seconds.\")\n",
        "print(\"Model is ready.\")\n",
        "model.save(Config.MODEL_PATH)\n",
        "print(\"Model saved : \", Config.MODEL_PATH)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}