ResNet50_FeatureEx.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "ResNet50_FeatureEx.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BIbvjN6j3y69",
        "outputId": "4a09f38b-833b-41d7-8583-da78e17a8e5d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Data Preprocess"
      ],
      "metadata": {
        "id": "fKZGYpZz4MP5"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mJWSHYEMpJEN"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import cv2\n",
        "import skimage\n",
        "from skimage.transform import resize\n",
        "\n",
        "import os\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torchvision\n",
        "from torchvision import models\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras.utils import to_categorical"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# !unzip -uq \"/content/drive/My Drive/ASL_Data.zip\"\n",
        "!unzip -o \"/content/drive/My Drive/ASL_Data.zip\" | awk 'BEGIN {ORS=\" \"} {if(NR%10==0)print \".\"}'"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QRaA5V6DHy4Y",
        "outputId": "6b1265e5-1e82-43aa-9144-b2f844a91e1c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            ". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . "
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "batch_size = 10\n",
        "imageSize = 76\n",
        "target_dims = (imageSize, imageSize, 3)\n",
        "num_classes = 29\n",
        "\n",
        "train_len = 87000\n",
        "train_dir = '/content/asl_alphabet_train/asl_alphabet_train/'"
      ],
      "metadata": {
        "id": "anCGjn9ILcVz"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def get_data(folder):\n",
        "    X = np.empty((train_len, imageSize, imageSize, 3), dtype=np.float32)\n",
        "    y = np.empty((train_len,), dtype=np.int64)\n",
        "    cnt = 0\n",
        "    for folderName in os.listdir(folder):\n",
        "        if not folderName.startswith('.'):\n",
        "            if folderName in ['A']:\n",
        "                label = 0\n",
        "            elif folderName in ['B']:\n",
        "                label = 1\n",
        "            elif folderName in ['C']:\n",
        "                label = 2\n",
        "            elif folderName in ['D']:\n",
        "                label = 3\n",
        "            elif folderName in ['E']:\n",
        "                label = 4\n",
        "            elif folderName in ['F']:\n",
        "                label = 5\n",
        "            elif folderName in ['G']:\n",
        "                label = 6\n",
        "            elif folderName in ['H']:\n",
        "                label = 7\n",
        "            elif folderName in ['I']:\n",
        "                label = 8\n",
        "            elif folderName in ['J']:\n",
        "                label = 9\n",
        "            elif folderName in ['K']:\n",
        "                label = 10\n",
        "            elif folderName in ['L']:\n",
        "                label = 11\n",
        "            elif folderName in ['M']:\n",
        "                label = 12\n",
        "            elif folderName in ['N']:\n",
        "                label = 13\n",
        "            elif folderName in ['O']:\n",
        "                label = 14\n",
        "            elif folderName in ['P']:\n",
        "                label = 15\n",
        "            elif folderName in ['Q']:\n",
        "                label = 16\n",
        "            elif folderName in ['R']:\n",
        "                label = 17\n",
        "            elif folderName in ['S']:\n",
        "                label = 18\n",
        "            elif folderName in ['T']:\n",
        "                label = 19\n",
        "            elif folderName in ['U']:\n",
        "                label = 20\n",
        "            elif folderName in ['V']:\n",
        "                label = 21\n",
        "            elif folderName in ['W']:\n",
        "                label = 22\n",
        "            elif folderName in ['X']:\n",
        "                label = 23\n",
        "            elif folderName in ['Y']:\n",
        "                label = 24\n",
        "            elif folderName in ['Z']:\n",
        "                label = 25\n",
        "            elif folderName in ['del']:\n",
        "                label = 26\n",
        "            elif folderName in ['nothing']:\n",
        "                label = 27\n",
        "            elif folderName in ['space']:\n",
        "                label = 28           \n",
        "            else:\n",
        "                label = 29\n",
        "            for image_filename in os.listdir(folder + folderName):\n",
        "                img_file = cv2.imread(folder + folderName + '/' + image_filename)\n",
        "                if img_file is not None:\n",
        "                    img_file = skimage.transform.resize(img_file, (imageSize, imageSize, 3))\n",
        "                    img_arr = np.asarray(img_file).reshape((-1, imageSize, imageSize, 3))\n",
        "                    \n",
        "                    X[cnt] = img_arr\n",
        "                    y[cnt] = label\n",
        "                    cnt += 1\n",
        "    return X,y"
      ],
      "metadata": {
        "id": "6BXtpAHyLoGl"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_train, y_train = get_data(train_dir)\n",
        "print(\"Images successfully imported...\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pkftXGpJLqAU",
        "outputId": "550a226e-e5c2-4399-d1e2-a081423cc468"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Images successfully imported...\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# train = pd.read_csv(\"/content/drive/MyDrive/archive/sign_mnist_train/sign_mnist_train.csv\")\n",
        "# test = pd.read_csv(\"/content/drive/MyDrive/archive/sign_mnist_test/sign_mnist_test.csv\")"
      ],
      "metadata": {
        "id": "0ccAVZKipK85"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(\"The shape of X_train is : \", X_train.shape)\n",
        "print(\"The shape of y_train is : \", y_train.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1azLL6sNqBzh",
        "outputId": "c0d8db20-0d22-4b46-9ed1-cd39a828a838"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The shape of X_train is :  (87000, 76, 76, 3)\n",
            "The shape of y_train is :  (87000,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.imshow(X_train[0])\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 268
        },
        "id": "phMp8nSJOR48",
        "outputId": "9de40270-7aaf-44ac-e479-f34551770d15"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X_train, X_val_test, y_train, y_val_test = train_test_split(X_train, y_train, test_size=0.4,random_state=42,stratify=y_train)\n",
        "X_val, X_test, y_val, y_test = train_test_split(X_val_test, y_val_test, test_size=0.5,random_state=42,stratify=y_val_test)"
      ],
      "metadata": {
        "id": "CGp9mEz5OfsO"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_cat_train = to_categorical(y_train,29)\n",
        "y_cat_val = to_categorical(y_val,29)\n",
        "y_cat_test = to_categorical(y_test,29)"
      ],
      "metadata": {
        "id": "z_T6UfbsOuti"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(X_train.shape)\n",
        "print(y_train.shape)\n",
        "print(X_val.shape)\n",
        "print(y_val.shape)\n",
        "print(X_test.shape)\n",
        "print(y_test.shape)\n",
        "print(y_cat_train.shape)\n",
        "print(y_cat_val.shape)\n",
        "print(y_cat_test.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QgiX6mPZPDYL",
        "outputId": "a30fffdf-36fd-423b-bbb3-0a8a07bce80f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(52200, 76, 76, 3)\n",
            "(52200,)\n",
            "(17400, 76, 76, 3)\n",
            "(17400,)\n",
            "(17400, 76, 76, 3)\n",
            "(17400,)\n",
            "(52200, 29)\n",
            "(17400, 29)\n",
            "(17400, 29)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(X_train[0].shape)\n",
        "\n",
        "image = cv2.resize(X_train[0], (299, 299))\n",
        "\n",
        "plt.imshow(image)\n",
        "print(y_train[0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 304
        },
        "id": "hUDaGbCVrThA",
        "outputId": "1a6c4d06-dc84-4b11-c249-188fd99c0a5b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(76, 76, 3)\n",
            "22\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "\n",
        "data_augmentation = keras.Sequential(\n",
        "    [layers.RandomFlip(\"horizontal\"), layers.RandomRotation(0.1),]\n",
        ")"
      ],
      "metadata": {
        "id": "bA1AGS9gA0s9"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.engine import training\n",
        "from keras.models import Sequential\n",
        "from keras.models import Model\n",
        "from keras.callbacks import ModelCheckpoint, LearningRateScheduler, EarlyStopping, ReduceLROnPlateau, TensorBoard\n",
        "from keras import optimizers, losses, activations, models\n",
        "from keras.layers import Convolution2D, Dense, Input, Flatten, Dropout, MaxPooling2D, BatchNormalization, GlobalAveragePooling2D, Concatenate\n",
        "from keras import applications"
      ],
      "metadata": {
        "id": "2xmnpGRqbZsw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 1st Extraction Layer"
      ],
      "metadata": {
        "id": "rlV0Y39YdD5N"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "base_model = tf.keras.applications.ResNet50(weights='imagenet', \n",
        "                                               include_top=False,\n",
        "                                               input_shape=(76, 76, 3))\n",
        "base_model.trainable = False\n",
        "resnet = Model(inputs=base_model.input, \n",
        "                  outputs=base_model.get_layer('conv2_block3_out').output)"
      ],
      "metadata": {
        "id": "zg4T_uJgiU02"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "base_model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BbC1VKn2diXr",
        "outputId": "1e20e858-8317-4ca0-c5be-a98ed3a263b2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"resnet50\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)           [(None, 76, 76, 3)]  0           []                               \n",
            "                                                                                                  \n",
            " conv1_pad (ZeroPadding2D)      (None, 82, 82, 3)    0           ['input_1[0][0]']                \n",
            "                                                                                                  \n",
            " conv1_conv (Conv2D)            (None, 38, 38, 64)   9472        ['conv1_pad[0][0]']              \n",
            "                                                                                                  \n",
            " conv1_bn (BatchNormalization)  (None, 38, 38, 64)   256         ['conv1_conv[0][0]']             \n",
            "                                                                                                  \n",
            " conv1_relu (Activation)        (None, 38, 38, 64)   0           ['conv1_bn[0][0]']               \n",
            "                                                                                                  \n",
            " pool1_pad (ZeroPadding2D)      (None, 40, 40, 64)   0           ['conv1_relu[0][0]']             \n",
            "                                                                                                  \n",
            " pool1_pool (MaxPooling2D)      (None, 19, 19, 64)   0           ['pool1_pad[0][0]']              \n",
            "                                                                                                  \n",
            " conv2_block1_1_conv (Conv2D)   (None, 19, 19, 64)   4160        ['pool1_pool[0][0]']             \n",
            "                                                                                                  \n",
            " conv2_block1_1_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block1_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block1_1_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block1_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block1_2_conv (Conv2D)   (None, 19, 19, 64)   36928       ['conv2_block1_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block1_2_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block1_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block1_2_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block1_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block1_0_conv (Conv2D)   (None, 19, 19, 256)  16640       ['pool1_pool[0][0]']             \n",
            "                                                                                                  \n",
            " conv2_block1_3_conv (Conv2D)   (None, 19, 19, 256)  16640       ['conv2_block1_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block1_0_bn (BatchNormal  (None, 19, 19, 256)  1024       ['conv2_block1_0_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block1_3_bn (BatchNormal  (None, 19, 19, 256)  1024       ['conv2_block1_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block1_add (Add)         (None, 19, 19, 256)  0           ['conv2_block1_0_bn[0][0]',      \n",
            "                                                                  'conv2_block1_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv2_block1_out (Activation)  (None, 19, 19, 256)  0           ['conv2_block1_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv2_block2_1_conv (Conv2D)   (None, 19, 19, 64)   16448       ['conv2_block1_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv2_block2_1_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block2_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block2_1_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block2_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block2_2_conv (Conv2D)   (None, 19, 19, 64)   36928       ['conv2_block2_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block2_2_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block2_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block2_2_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block2_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block2_3_conv (Conv2D)   (None, 19, 19, 256)  16640       ['conv2_block2_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block2_3_bn (BatchNormal  (None, 19, 19, 256)  1024       ['conv2_block2_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block2_add (Add)         (None, 19, 19, 256)  0           ['conv2_block1_out[0][0]',       \n",
            "                                                                  'conv2_block2_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv2_block2_out (Activation)  (None, 19, 19, 256)  0           ['conv2_block2_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv2_block3_1_conv (Conv2D)   (None, 19, 19, 64)   16448       ['conv2_block2_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv2_block3_1_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block3_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block3_1_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block3_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block3_2_conv (Conv2D)   (None, 19, 19, 64)   36928       ['conv2_block3_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block3_2_bn (BatchNormal  (None, 19, 19, 64)  256         ['conv2_block3_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block3_2_relu (Activatio  (None, 19, 19, 64)  0           ['conv2_block3_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv2_block3_3_conv (Conv2D)   (None, 19, 19, 256)  16640       ['conv2_block3_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv2_block3_3_bn (BatchNormal  (None, 19, 19, 256)  1024       ['conv2_block3_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv2_block3_add (Add)         (None, 19, 19, 256)  0           ['conv2_block2_out[0][0]',       \n",
            "                                                                  'conv2_block3_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv2_block3_out (Activation)  (None, 19, 19, 256)  0           ['conv2_block3_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block1_1_conv (Conv2D)   (None, 10, 10, 128)  32896       ['conv2_block3_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block1_1_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block1_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block1_1_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block1_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block1_2_conv (Conv2D)   (None, 10, 10, 128)  147584      ['conv3_block1_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block1_2_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block1_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block1_2_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block1_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block1_0_conv (Conv2D)   (None, 10, 10, 512)  131584      ['conv2_block3_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block1_3_conv (Conv2D)   (None, 10, 10, 512)  66048       ['conv3_block1_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block1_0_bn (BatchNormal  (None, 10, 10, 512)  2048       ['conv3_block1_0_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block1_3_bn (BatchNormal  (None, 10, 10, 512)  2048       ['conv3_block1_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block1_add (Add)         (None, 10, 10, 512)  0           ['conv3_block1_0_bn[0][0]',      \n",
            "                                                                  'conv3_block1_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv3_block1_out (Activation)  (None, 10, 10, 512)  0           ['conv3_block1_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block2_1_conv (Conv2D)   (None, 10, 10, 128)  65664       ['conv3_block1_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block2_1_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block2_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block2_1_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block2_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block2_2_conv (Conv2D)   (None, 10, 10, 128)  147584      ['conv3_block2_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block2_2_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block2_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block2_2_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block2_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block2_3_conv (Conv2D)   (None, 10, 10, 512)  66048       ['conv3_block2_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block2_3_bn (BatchNormal  (None, 10, 10, 512)  2048       ['conv3_block2_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block2_add (Add)         (None, 10, 10, 512)  0           ['conv3_block1_out[0][0]',       \n",
            "                                                                  'conv3_block2_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv3_block2_out (Activation)  (None, 10, 10, 512)  0           ['conv3_block2_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block3_1_conv (Conv2D)   (None, 10, 10, 128)  65664       ['conv3_block2_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block3_1_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block3_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block3_1_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block3_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block3_2_conv (Conv2D)   (None, 10, 10, 128)  147584      ['conv3_block3_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block3_2_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block3_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block3_2_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block3_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block3_3_conv (Conv2D)   (None, 10, 10, 512)  66048       ['conv3_block3_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block3_3_bn (BatchNormal  (None, 10, 10, 512)  2048       ['conv3_block3_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block3_add (Add)         (None, 10, 10, 512)  0           ['conv3_block2_out[0][0]',       \n",
            "                                                                  'conv3_block3_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv3_block3_out (Activation)  (None, 10, 10, 512)  0           ['conv3_block3_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block4_1_conv (Conv2D)   (None, 10, 10, 128)  65664       ['conv3_block3_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv3_block4_1_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block4_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block4_1_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block4_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block4_2_conv (Conv2D)   (None, 10, 10, 128)  147584      ['conv3_block4_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block4_2_bn (BatchNormal  (None, 10, 10, 128)  512        ['conv3_block4_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block4_2_relu (Activatio  (None, 10, 10, 128)  0          ['conv3_block4_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv3_block4_3_conv (Conv2D)   (None, 10, 10, 512)  66048       ['conv3_block4_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv3_block4_3_bn (BatchNormal  (None, 10, 10, 512)  2048       ['conv3_block4_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv3_block4_add (Add)         (None, 10, 10, 512)  0           ['conv3_block3_out[0][0]',       \n",
            "                                                                  'conv3_block4_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv3_block4_out (Activation)  (None, 10, 10, 512)  0           ['conv3_block4_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block1_1_conv (Conv2D)   (None, 5, 5, 256)    131328      ['conv3_block4_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block1_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block1_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block1_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block1_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block1_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block1_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block1_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block1_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block1_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block1_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block1_0_conv (Conv2D)   (None, 5, 5, 1024)   525312      ['conv3_block4_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block1_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block1_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block1_0_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block1_0_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block1_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block1_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block1_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block1_0_bn[0][0]',      \n",
            "                                                                  'conv4_block1_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block1_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block1_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block2_1_conv (Conv2D)   (None, 5, 5, 256)    262400      ['conv4_block1_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block2_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block2_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block2_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block2_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block2_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block2_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block2_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block2_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block2_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block2_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block2_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block2_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block2_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block2_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block2_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block1_out[0][0]',       \n",
            "                                                                  'conv4_block2_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block2_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block2_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block3_1_conv (Conv2D)   (None, 5, 5, 256)    262400      ['conv4_block2_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block3_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block3_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block3_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block3_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block3_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block3_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block3_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block3_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block3_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block3_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block3_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block3_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block3_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block3_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block3_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block2_out[0][0]',       \n",
            "                                                                  'conv4_block3_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block3_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block3_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block4_1_conv (Conv2D)   (None, 5, 5, 256)    262400      ['conv4_block3_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block4_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block4_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block4_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block4_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block4_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block4_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block4_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block4_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block4_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block4_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block4_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block4_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block4_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block4_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block4_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block3_out[0][0]',       \n",
            "                                                                  'conv4_block4_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block4_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block4_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block5_1_conv (Conv2D)   (None, 5, 5, 256)    262400      ['conv4_block4_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block5_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block5_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block5_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block5_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block5_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block5_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block5_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block5_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block5_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block5_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block5_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block5_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block5_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block5_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block5_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block4_out[0][0]',       \n",
            "                                                                  'conv4_block5_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block5_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block5_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block6_1_conv (Conv2D)   (None, 5, 5, 256)    262400      ['conv4_block5_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv4_block6_1_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block6_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block6_1_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block6_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block6_2_conv (Conv2D)   (None, 5, 5, 256)    590080      ['conv4_block6_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block6_2_bn (BatchNormal  (None, 5, 5, 256)   1024        ['conv4_block6_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block6_2_relu (Activatio  (None, 5, 5, 256)   0           ['conv4_block6_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv4_block6_3_conv (Conv2D)   (None, 5, 5, 1024)   263168      ['conv4_block6_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv4_block6_3_bn (BatchNormal  (None, 5, 5, 1024)  4096        ['conv4_block6_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv4_block6_add (Add)         (None, 5, 5, 1024)   0           ['conv4_block5_out[0][0]',       \n",
            "                                                                  'conv4_block6_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv4_block6_out (Activation)  (None, 5, 5, 1024)   0           ['conv4_block6_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block1_1_conv (Conv2D)   (None, 3, 3, 512)    524800      ['conv4_block6_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block1_1_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block1_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block1_1_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block1_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block1_2_conv (Conv2D)   (None, 3, 3, 512)    2359808     ['conv5_block1_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block1_2_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block1_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block1_2_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block1_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block1_0_conv (Conv2D)   (None, 3, 3, 2048)   2099200     ['conv4_block6_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block1_3_conv (Conv2D)   (None, 3, 3, 2048)   1050624     ['conv5_block1_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block1_0_bn (BatchNormal  (None, 3, 3, 2048)  8192        ['conv5_block1_0_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block1_3_bn (BatchNormal  (None, 3, 3, 2048)  8192        ['conv5_block1_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block1_add (Add)         (None, 3, 3, 2048)   0           ['conv5_block1_0_bn[0][0]',      \n",
            "                                                                  'conv5_block1_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv5_block1_out (Activation)  (None, 3, 3, 2048)   0           ['conv5_block1_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block2_1_conv (Conv2D)   (None, 3, 3, 512)    1049088     ['conv5_block1_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block2_1_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block2_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block2_1_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block2_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block2_2_conv (Conv2D)   (None, 3, 3, 512)    2359808     ['conv5_block2_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block2_2_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block2_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block2_2_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block2_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block2_3_conv (Conv2D)   (None, 3, 3, 2048)   1050624     ['conv5_block2_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block2_3_bn (BatchNormal  (None, 3, 3, 2048)  8192        ['conv5_block2_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block2_add (Add)         (None, 3, 3, 2048)   0           ['conv5_block1_out[0][0]',       \n",
            "                                                                  'conv5_block2_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv5_block2_out (Activation)  (None, 3, 3, 2048)   0           ['conv5_block2_add[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block3_1_conv (Conv2D)   (None, 3, 3, 512)    1049088     ['conv5_block2_out[0][0]']       \n",
            "                                                                                                  \n",
            " conv5_block3_1_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block3_1_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block3_1_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block3_1_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block3_2_conv (Conv2D)   (None, 3, 3, 512)    2359808     ['conv5_block3_1_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block3_2_bn (BatchNormal  (None, 3, 3, 512)   2048        ['conv5_block3_2_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block3_2_relu (Activatio  (None, 3, 3, 512)   0           ['conv5_block3_2_bn[0][0]']      \n",
            " n)                                                                                               \n",
            "                                                                                                  \n",
            " conv5_block3_3_conv (Conv2D)   (None, 3, 3, 2048)   1050624     ['conv5_block3_2_relu[0][0]']    \n",
            "                                                                                                  \n",
            " conv5_block3_3_bn (BatchNormal  (None, 3, 3, 2048)  8192        ['conv5_block3_3_conv[0][0]']    \n",
            " ization)                                                                                         \n",
            "                                                                                                  \n",
            " conv5_block3_add (Add)         (None, 3, 3, 2048)   0           ['conv5_block2_out[0][0]',       \n",
            "                                                                  'conv5_block3_3_bn[0][0]']      \n",
            "                                                                                                  \n",
            " conv5_block3_out (Activation)  (None, 3, 3, 2048)   0           ['conv5_block3_add[0][0]']       \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 23,587,712\n",
            "Trainable params: 0\n",
            "Non-trainable params: 23,587,712\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "nclass = 29\n",
        "\n",
        "feature_extraction = Sequential()\n",
        "feature_extraction.add(resnet)\n",
        "feature_extraction.add(Flatten())\n",
        "feature_extraction.add(Dense(nclass, activation='softmax'))\n",
        "\n",
        "feature_extraction.compile(loss='categorical_crossentropy', \n",
        "                           optimizer=keras.optimizers.Adam(learning_rate=0.01), \n",
        "                           metrics=['accuracy'])\n",
        "\n",
        "feature_extraction.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-i5-ycRwSBti",
        "outputId": "c8190b38-b4c8-417e-98ff-96bd41fcfe76"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " model (Functional)          (None, 19, 19, 256)       229760    \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 92416)             0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 29)                2680093   \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 2,909,853\n",
            "Trainable params: 2,680,093\n",
            "Non-trainable params: 229,760\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "file_path=\"/content/best_model.h5\"\n",
        "\n",
        "# checkpoint = ModelCheckpoint(file_path, monitor='acc', verbose=1, save_best_only=True, mode='max')\n",
        "mc = ModelCheckpoint(file_path, monitor='val_loss', verbose=1, mode='min', save_best_only=True)\n",
        "\n",
        "# early = EarlyStopping(monitor=\"acc\", mode=\"max\", patience=15)\n",
        "es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10)\n",
        "\n",
        "callbacks_list = [es, mc] #early\n",
        "\n",
        "feature_extraction.fit(x=X_train,\n",
        "          y=y_cat_train,\n",
        "          validation_data=(X_val, y_cat_val),\n",
        "          batch_size=32,\n",
        "          epochs=30, \n",
        "          shuffle=True, \n",
        "          verbose=2,\n",
        "          callbacks=callbacks_list)\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lcJfSR2lrMb8",
        "outputId": "7f7700b5-efb1-4919-daf1-961069ab92e6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "\n",
            "Epoch 1: val_loss improved from inf to 1007.35175, saving model to /content/best_model.h5\n",
            "1632/1632 - 15s - loss: 1292.5649 - accuracy: 0.0431 - val_loss: 1007.3517 - val_accuracy: 0.0471 - 15s/epoch - 9ms/step\n",
            "Epoch 2/30\n",
            "\n",
            "Epoch 2: val_loss improved from 1007.35175 to 826.10864, saving model to /content/best_model.h5\n",
            "1632/1632 - 11s - loss: 964.2982 - accuracy: 0.0728 - val_loss: 826.1086 - val_accuracy: 0.0889 - 11s/epoch - 7ms/step\n",
            "Epoch 3/30\n",
            "\n",
            "Epoch 3: val_loss improved from 826.10864 to 598.07733, saving model to /content/best_model.h5\n",
            "1632/1632 - 11s - loss: 790.1287 - accuracy: 0.1129 - val_loss: 598.0773 - val_accuracy: 0.1250 - 11s/epoch - 7ms/step\n",
            "Epoch 4/30\n",
            "\n",
            "Epoch 4: val_loss improved from 598.07733 to 505.22821, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 665.8110 - accuracy: 0.1544 - val_loss: 505.2282 - val_accuracy: 0.2192 - 12s/epoch - 7ms/step\n",
            "Epoch 5/30\n",
            "\n",
            "Epoch 5: val_loss did not improve from 505.22821\n",
            "1632/1632 - 12s - loss: 621.5911 - accuracy: 0.1762 - val_loss: 573.3712 - val_accuracy: 0.1654 - 12s/epoch - 7ms/step\n",
            "Epoch 6/30\n",
            "\n",
            "Epoch 6: val_loss improved from 505.22821 to 377.01419, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 541.0337 - accuracy: 0.2135 - val_loss: 377.0142 - val_accuracy: 0.3362 - 12s/epoch - 7ms/step\n",
            "Epoch 7/30\n",
            "\n",
            "Epoch 7: val_loss did not improve from 377.01419\n",
            "1632/1632 - 12s - loss: 497.7939 - accuracy: 0.2325 - val_loss: 420.4691 - val_accuracy: 0.2151 - 12s/epoch - 7ms/step\n",
            "Epoch 8/30\n",
            "\n",
            "Epoch 8: val_loss did not improve from 377.01419\n",
            "1632/1632 - 12s - loss: 472.2114 - accuracy: 0.2519 - val_loss: 449.8826 - val_accuracy: 0.2045 - 12s/epoch - 7ms/step\n",
            "Epoch 9/30\n",
            "\n",
            "Epoch 9: val_loss did not improve from 377.01419\n",
            "1632/1632 - 12s - loss: 450.6257 - accuracy: 0.2690 - val_loss: 405.2970 - val_accuracy: 0.1914 - 12s/epoch - 7ms/step\n",
            "Epoch 10/30\n",
            "\n",
            "Epoch 10: val_loss improved from 377.01419 to 371.37854, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 364.1034 - accuracy: 0.3075 - val_loss: 371.3785 - val_accuracy: 0.2775 - 12s/epoch - 7ms/step\n",
            "Epoch 11/30\n",
            "\n",
            "Epoch 11: val_loss did not improve from 371.37854\n",
            "1632/1632 - 12s - loss: 373.8858 - accuracy: 0.3136 - val_loss: 513.7838 - val_accuracy: 0.2977 - 12s/epoch - 7ms/step\n",
            "Epoch 12/30\n",
            "\n",
            "Epoch 12: val_loss did not improve from 371.37854\n",
            "1632/1632 - 12s - loss: 340.8174 - accuracy: 0.3426 - val_loss: 381.7517 - val_accuracy: 0.2984 - 12s/epoch - 7ms/step\n",
            "Epoch 13/30\n",
            "\n",
            "Epoch 13: val_loss improved from 371.37854 to 358.69141, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 339.8284 - accuracy: 0.3488 - val_loss: 358.6914 - val_accuracy: 0.2657 - 12s/epoch - 7ms/step\n",
            "Epoch 14/30\n",
            "\n",
            "Epoch 14: val_loss improved from 358.69141 to 255.11508, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 319.4976 - accuracy: 0.3670 - val_loss: 255.1151 - val_accuracy: 0.3118 - 12s/epoch - 7ms/step\n",
            "Epoch 15/30\n",
            "\n",
            "Epoch 15: val_loss did not improve from 255.11508\n",
            "1632/1632 - 12s - loss: 312.5603 - accuracy: 0.3761 - val_loss: 321.1813 - val_accuracy: 0.3175 - 12s/epoch - 7ms/step\n",
            "Epoch 16/30\n",
            "\n",
            "Epoch 16: val_loss did not improve from 255.11508\n",
            "1632/1632 - 11s - loss: 287.6311 - accuracy: 0.3959 - val_loss: 374.7293 - val_accuracy: 0.4040 - 11s/epoch - 7ms/step\n",
            "Epoch 17/30\n",
            "\n",
            "Epoch 17: val_loss improved from 255.11508 to 227.02774, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 292.8978 - accuracy: 0.3971 - val_loss: 227.0277 - val_accuracy: 0.4732 - 12s/epoch - 7ms/step\n",
            "Epoch 18/30\n",
            "\n",
            "Epoch 18: val_loss did not improve from 227.02774\n",
            "1632/1632 - 12s - loss: 262.9327 - accuracy: 0.4184 - val_loss: 797.6916 - val_accuracy: 0.2539 - 12s/epoch - 7ms/step\n",
            "Epoch 19/30\n",
            "\n",
            "Epoch 19: val_loss did not improve from 227.02774\n",
            "1632/1632 - 12s - loss: 268.7768 - accuracy: 0.4189 - val_loss: 349.3305 - val_accuracy: 0.3501 - 12s/epoch - 7ms/step\n",
            "Epoch 20/30\n",
            "\n",
            "Epoch 20: val_loss improved from 227.02774 to 171.48560, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 267.3575 - accuracy: 0.4318 - val_loss: 171.4856 - val_accuracy: 0.4668 - 12s/epoch - 7ms/step\n",
            "Epoch 21/30\n",
            "\n",
            "Epoch 21: val_loss did not improve from 171.48560\n",
            "1632/1632 - 12s - loss: 243.8746 - accuracy: 0.4479 - val_loss: 196.7635 - val_accuracy: 0.5002 - 12s/epoch - 7ms/step\n",
            "Epoch 22/30\n",
            "\n",
            "Epoch 22: val_loss improved from 171.48560 to 123.17273, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 256.1548 - accuracy: 0.4483 - val_loss: 123.1727 - val_accuracy: 0.4912 - 12s/epoch - 7ms/step\n",
            "Epoch 23/30\n",
            "\n",
            "Epoch 23: val_loss did not improve from 123.17273\n",
            "1632/1632 - 11s - loss: 231.4942 - accuracy: 0.4705 - val_loss: 248.4945 - val_accuracy: 0.3558 - 11s/epoch - 7ms/step\n",
            "Epoch 24/30\n",
            "\n",
            "Epoch 24: val_loss did not improve from 123.17273\n",
            "1632/1632 - 12s - loss: 246.7791 - accuracy: 0.4643 - val_loss: 145.1610 - val_accuracy: 0.5094 - 12s/epoch - 7ms/step\n",
            "Epoch 25/30\n",
            "\n",
            "Epoch 25: val_loss did not improve from 123.17273\n",
            "1632/1632 - 12s - loss: 228.6753 - accuracy: 0.4792 - val_loss: 332.1891 - val_accuracy: 0.3137 - 12s/epoch - 7ms/step\n",
            "Epoch 26/30\n",
            "\n",
            "Epoch 26: val_loss did not improve from 123.17273\n",
            "1632/1632 - 12s - loss: 229.4521 - accuracy: 0.4861 - val_loss: 179.6752 - val_accuracy: 0.4764 - 12s/epoch - 7ms/step\n",
            "Epoch 27/30\n",
            "\n",
            "Epoch 27: val_loss improved from 123.17273 to 76.04976, saving model to /content/best_model.h5\n",
            "1632/1632 - 12s - loss: 218.3215 - accuracy: 0.4937 - val_loss: 76.0498 - val_accuracy: 0.6114 - 12s/epoch - 7ms/step\n",
            "Epoch 28/30\n",
            "\n",
            "Epoch 28: val_loss did not improve from 76.04976\n",
            "1632/1632 - 11s - loss: 218.2867 - accuracy: 0.4959 - val_loss: 400.8872 - val_accuracy: 0.3788 - 11s/epoch - 7ms/step\n",
            "Epoch 29/30\n",
            "\n",
            "Epoch 29: val_loss did not improve from 76.04976\n",
            "1632/1632 - 12s - loss: 218.4240 - accuracy: 0.5041 - val_loss: 193.6999 - val_accuracy: 0.4620 - 12s/epoch - 7ms/step\n",
            "Epoch 30/30\n",
            "\n",
            "Epoch 30: val_loss did not improve from 76.04976\n",
            "1632/1632 - 11s - loss: 201.8298 - accuracy: 0.5155 - val_loss: 94.6858 - val_accuracy: 0.5728 - 11s/epoch - 7ms/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fb4de1fea10>"
            ]
          },
          "metadata": {},
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "resnet_fe1 = [1007.3517, 826.1086, 598.0773, 505.2282, 573.3712, 377.0142, 420.4691, 449.8826, 405.2970, 371.3785, 513.7838, 381.7517, 358.6914, 255.1151, 321.1813, 374.7293, 227.0277, 797.6916, 349.3305, 171.4856, 196.7635, 123.1727, 248.4945, 145.1610, 332.1891, 179.6752, 76.0498, 400.8872, 193.6999, 94.6858]"
      ],
      "metadata": {
        "id": "THYedIilxI_Z"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "score = feature_extraction.evaluate(x = X_test, y = y_cat_test, verbose = 1)\n",
        "print('Accuracy for test images:', round(score[1]*100, 3), '%')"
      ],
      "metadata": {
        "id": "WhFxQwtWe5Ci",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "66eee176-fbb1-4db3-bbf8-d0effb3a80a9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "544/544 [==============================] - 3s 6ms/step - loss: 94.6993 - accuracy: 0.5751\n",
            "Accuracy for test images: 57.511 %\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 2nd Extraction Layer"
      ],
      "metadata": {
        "id": "K9nZfV1AdPLM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "resnet = Model(inputs=base_model.input, \n",
        "                  outputs=base_model.get_layer('conv3_block4_out').output)"
      ],
      "metadata": {
        "id": "jA0zHEXadPLP"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "nclass = 29\n",
        "\n",
        "feature_extraction = Sequential()\n",
        "feature_extraction.add(resnet)\n",
        "feature_extraction.add(Flatten())\n",
        "feature_extraction.add(Dense(nclass, activation='softmax'))\n",
        "\n",
        "feature_extraction.compile(loss='categorical_crossentropy', \n",
        "                           optimizer=keras.optimizers.Adam(), \n",
        "                           metrics=['accuracy'])\n",
        "\n",
        "feature_extraction.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "79f2f27b-d1c5-4578-efb4-1ccc1ff7a545",
        "id": "nq7ZF8etdPLS"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " model_1 (Functional)        (None, 10, 10, 512)       1460096   \n",
            "                                                                 \n",
            " flatten_1 (Flatten)         (None, 51200)             0         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 29)                1484829   \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 2,944,925\n",
            "Trainable params: 1,484,829\n",
            "Non-trainable params: 1,460,096\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "file_path=\"/content/best_model.h5\"\n",
        "\n",
        "# checkpoint = ModelCheckpoint(file_path, monitor='acc', verbose=1, save_best_only=True, mode='max')\n",
        "mc = ModelCheckpoint(file_path, monitor='val_loss', verbose=1, mode='min', save_best_only=True)\n",
        "\n",
        "# early = EarlyStopping(monitor=\"acc\", mode=\"max\", patience=15)\n",
        "es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10)\n",
        "\n",
        "callbacks_list = [es, mc] #early\n",
        "\n",
        "feature_extraction.fit(x=X_train,\n",
        "          y=y_cat_train,\n",
        "          validation_data=(X_val, y_cat_val),\n",
        "          batch_size=32,\n",
        "          epochs=30, \n",
        "          shuffle=True, \n",
        "          verbose=2,\n",
        "          callbacks=callbacks_list)\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3f9e687a-a649-423b-bfbc-85473e27c55c",
        "id": "yDX1YdXHdPLU"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "\n",
            "Epoch 1: val_loss improved from inf to 11.32158, saving model to /content/best_model.h5\n",
            "1632/1632 - 20s - loss: 23.4782 - accuracy: 0.1637 - val_loss: 11.3216 - val_accuracy: 0.2676 - 20s/epoch - 12ms/step\n",
            "Epoch 2/30\n",
            "\n",
            "Epoch 2: val_loss improved from 11.32158 to 10.63105, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 13.0813 - accuracy: 0.2992 - val_loss: 10.6311 - val_accuracy: 0.2875 - 16s/epoch - 10ms/step\n",
            "Epoch 3/30\n",
            "\n",
            "Epoch 3: val_loss improved from 10.63105 to 9.45678, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 11.1282 - accuracy: 0.3607 - val_loss: 9.4568 - val_accuracy: 0.3700 - 16s/epoch - 10ms/step\n",
            "Epoch 4/30\n",
            "\n",
            "Epoch 4: val_loss did not improve from 9.45678\n",
            "1632/1632 - 16s - loss: 10.3903 - accuracy: 0.4118 - val_loss: 11.5367 - val_accuracy: 0.3822 - 16s/epoch - 10ms/step\n",
            "Epoch 5/30\n",
            "\n",
            "Epoch 5: val_loss improved from 9.45678 to 4.93211, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 9.1614 - accuracy: 0.4498 - val_loss: 4.9321 - val_accuracy: 0.4975 - 16s/epoch - 10ms/step\n",
            "Epoch 6/30\n",
            "\n",
            "Epoch 6: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 8.8826 - accuracy: 0.4754 - val_loss: 10.2046 - val_accuracy: 0.4379 - 16s/epoch - 10ms/step\n",
            "Epoch 7/30\n",
            "\n",
            "Epoch 7: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 8.9594 - accuracy: 0.4889 - val_loss: 6.2379 - val_accuracy: 0.5410 - 16s/epoch - 10ms/step\n",
            "Epoch 8/30\n",
            "\n",
            "Epoch 8: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 7.8871 - accuracy: 0.5219 - val_loss: 10.8437 - val_accuracy: 0.4534 - 16s/epoch - 10ms/step\n",
            "Epoch 9/30\n",
            "\n",
            "Epoch 9: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 8.0388 - accuracy: 0.5292 - val_loss: 8.9360 - val_accuracy: 0.4648 - 16s/epoch - 10ms/step\n",
            "Epoch 10/30\n",
            "\n",
            "Epoch 10: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 7.5505 - accuracy: 0.5511 - val_loss: 7.2677 - val_accuracy: 0.5410 - 16s/epoch - 10ms/step\n",
            "Epoch 11/30\n",
            "\n",
            "Epoch 11: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 7.2803 - accuracy: 0.5661 - val_loss: 9.7766 - val_accuracy: 0.5353 - 16s/epoch - 10ms/step\n",
            "Epoch 12/30\n",
            "\n",
            "Epoch 12: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 7.2683 - accuracy: 0.5764 - val_loss: 9.2243 - val_accuracy: 0.5116 - 16s/epoch - 10ms/step\n",
            "Epoch 13/30\n",
            "\n",
            "Epoch 13: val_loss did not improve from 4.93211\n",
            "1632/1632 - 16s - loss: 7.1531 - accuracy: 0.5854 - val_loss: 5.8110 - val_accuracy: 0.5630 - 16s/epoch - 10ms/step\n",
            "Epoch 14/30\n",
            "\n",
            "Epoch 14: val_loss improved from 4.93211 to 4.16894, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 6.7147 - accuracy: 0.6007 - val_loss: 4.1689 - val_accuracy: 0.6322 - 16s/epoch - 10ms/step\n",
            "Epoch 15/30\n",
            "\n",
            "Epoch 15: val_loss did not improve from 4.16894\n",
            "1632/1632 - 16s - loss: 6.6002 - accuracy: 0.6081 - val_loss: 10.9864 - val_accuracy: 0.5532 - 16s/epoch - 10ms/step\n",
            "Epoch 16/30\n",
            "\n",
            "Epoch 16: val_loss did not improve from 4.16894\n",
            "1632/1632 - 16s - loss: 6.4127 - accuracy: 0.6154 - val_loss: 6.3322 - val_accuracy: 0.5865 - 16s/epoch - 10ms/step\n",
            "Epoch 17/30\n",
            "\n",
            "Epoch 17: val_loss did not improve from 4.16894\n",
            "1632/1632 - 16s - loss: 6.5896 - accuracy: 0.6248 - val_loss: 4.6960 - val_accuracy: 0.6325 - 16s/epoch - 10ms/step\n",
            "Epoch 18/30\n",
            "\n",
            "Epoch 18: val_loss did not improve from 4.16894\n",
            "1632/1632 - 16s - loss: 6.6900 - accuracy: 0.6275 - val_loss: 6.3588 - val_accuracy: 0.6015 - 16s/epoch - 10ms/step\n",
            "Epoch 19/30\n",
            "\n",
            "Epoch 19: val_loss improved from 4.16894 to 4.14352, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 6.2033 - accuracy: 0.6408 - val_loss: 4.1435 - val_accuracy: 0.6709 - 16s/epoch - 10ms/step\n",
            "Epoch 20/30\n",
            "\n",
            "Epoch 20: val_loss did not improve from 4.14352\n",
            "1632/1632 - 16s - loss: 6.5570 - accuracy: 0.6380 - val_loss: 6.6157 - val_accuracy: 0.6112 - 16s/epoch - 10ms/step\n",
            "Epoch 21/30\n",
            "\n",
            "Epoch 21: val_loss did not improve from 4.14352\n",
            "1632/1632 - 16s - loss: 6.1058 - accuracy: 0.6527 - val_loss: 4.6213 - val_accuracy: 0.6708 - 16s/epoch - 10ms/step\n",
            "Epoch 22/30\n",
            "\n",
            "Epoch 22: val_loss did not improve from 4.14352\n",
            "1632/1632 - 16s - loss: 5.9827 - accuracy: 0.6585 - val_loss: 7.9544 - val_accuracy: 0.6328 - 16s/epoch - 10ms/step\n",
            "Epoch 23/30\n",
            "\n",
            "Epoch 23: val_loss improved from 4.14352 to 3.19416, saving model to /content/best_model.h5\n",
            "1632/1632 - 16s - loss: 5.8094 - accuracy: 0.6664 - val_loss: 3.1942 - val_accuracy: 0.7230 - 16s/epoch - 10ms/step\n",
            "Epoch 24/30\n",
            "\n",
            "Epoch 24: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 6.3498 - accuracy: 0.6598 - val_loss: 3.8277 - val_accuracy: 0.7072 - 16s/epoch - 10ms/step\n",
            "Epoch 25/30\n",
            "\n",
            "Epoch 25: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.8130 - accuracy: 0.6732 - val_loss: 6.9936 - val_accuracy: 0.6535 - 16s/epoch - 10ms/step\n",
            "Epoch 26/30\n",
            "\n",
            "Epoch 26: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.6188 - accuracy: 0.6793 - val_loss: 10.2872 - val_accuracy: 0.6022 - 16s/epoch - 10ms/step\n",
            "Epoch 27/30\n",
            "\n",
            "Epoch 27: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.4162 - accuracy: 0.6872 - val_loss: 3.6271 - val_accuracy: 0.7201 - 16s/epoch - 10ms/step\n",
            "Epoch 28/30\n",
            "\n",
            "Epoch 28: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.5658 - accuracy: 0.6902 - val_loss: 5.1532 - val_accuracy: 0.6794 - 16s/epoch - 10ms/step\n",
            "Epoch 29/30\n",
            "\n",
            "Epoch 29: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.6382 - accuracy: 0.6918 - val_loss: 5.8224 - val_accuracy: 0.7117 - 16s/epoch - 10ms/step\n",
            "Epoch 30/30\n",
            "\n",
            "Epoch 30: val_loss did not improve from 3.19416\n",
            "1632/1632 - 16s - loss: 5.5120 - accuracy: 0.6943 - val_loss: 6.7240 - val_accuracy: 0.6676 - 16s/epoch - 10ms/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fb4b8279b10>"
            ]
          },
          "metadata": {},
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "resnet_fe2 = [11.3216, 10.6311, 9.4568, 11.5367, 4.9321, 10.2046, 6.2379, 10.8437, 8.9360, 7.2677, 9.7766, 9.2243, 5.8110, 4.1689, 10.9864, 6.3322, 4.6960, 6.3588, 4.1435, 6.6157, 4.6213, 7.9544, 3.1942, 3.8277, 6.9936, 10.2872, 3.6271, 5.1532, 5.8224, 6.7240]"
      ],
      "metadata": {
        "id": "yx2ESsHUx7Rz"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "score = feature_extraction.evaluate(x = X_test, y = y_cat_test, verbose = 1)\n",
        "print('Accuracy for test images:', round(score[1]*100, 3), '%')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4d3db8d8-5d1f-47d0-d4ce-e37d46ac2e14",
        "id": "AvLSMWBUdPLW"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "544/544 [==============================] - 4s 8ms/step - loss: 6.8237 - accuracy: 0.6648\n",
            "Accuracy for test images: 66.477 %\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 3rd Extraction Layer"
      ],
      "metadata": {
        "id": "aavU_LW-eMl7"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "resnet = Model(inputs=base_model.input, \n",
        "                  outputs=base_model.get_layer('conv5_block3_out').output)"
      ],
      "metadata": {
        "id": "60WQaAa4eMl_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "nclass = 29\n",
        "\n",
        "feature_extraction = Sequential()\n",
        "feature_extraction.add(resnet)\n",
        "feature_extraction.add(Flatten())\n",
        "feature_extraction.add(Dense(nclass, activation='softmax'))\n",
        "\n",
        "feature_extraction.compile(loss='categorical_crossentropy', \n",
        "                           optimizer=keras.optimizers.Adam(), \n",
        "                           metrics=['accuracy'])\n",
        "\n",
        "feature_extraction.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d6977f56-ddb4-481f-ca7f-fb1b300ff02c",
        "id": "uBYtQQ9feMmB"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_3\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " model_2 (Functional)        (None, 3, 3, 2048)        23587712  \n",
            "                                                                 \n",
            " flatten_2 (Flatten)         (None, 18432)             0         \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 29)                534557    \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 24,122,269\n",
            "Trainable params: 534,557\n",
            "Non-trainable params: 23,587,712\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "file_path=\"/content/best_model.h5\"\n",
        "\n",
        "# checkpoint = ModelCheckpoint(file_path, monitor='acc', verbose=1, save_best_only=True, mode='max')\n",
        "mc = ModelCheckpoint(file_path, monitor='val_loss', verbose=1, mode='min', save_best_only=True)\n",
        "\n",
        "# early = EarlyStopping(monitor=\"acc\", mode=\"max\", patience=15)\n",
        "es = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10)\n",
        "\n",
        "callbacks_list = [es, mc] #early\n",
        "\n",
        "feature_extraction.fit(x=X_train,\n",
        "          y=y_cat_train,\n",
        "          validation_data=(X_val, y_cat_val),\n",
        "          batch_size=32,\n",
        "          epochs=30, \n",
        "          shuffle=True, \n",
        "          verbose=2,\n",
        "          callbacks=callbacks_list)\n",
        "\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ab190ba5-69d3-4b38-c264-c1966b6c1ed4",
        "id": "C6FAGcNkeMmC"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "\n",
            "Epoch 1: val_loss improved from inf to 1.86604, saving model to /content/best_model.h5\n",
            "1632/1632 - 40s - loss: 2.3136 - accuracy: 0.3399 - val_loss: 1.8660 - val_accuracy: 0.4459 - 40s/epoch - 25ms/step\n",
            "Epoch 2/30\n",
            "\n",
            "Epoch 2: val_loss improved from 1.86604 to 1.55871, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.6994 - accuracy: 0.5087 - val_loss: 1.5587 - val_accuracy: 0.5503 - 36s/epoch - 22ms/step\n",
            "Epoch 3/30\n",
            "\n",
            "Epoch 3: val_loss improved from 1.55871 to 1.46873, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.4620 - accuracy: 0.5728 - val_loss: 1.4687 - val_accuracy: 0.5553 - 36s/epoch - 22ms/step\n",
            "Epoch 4/30\n",
            "\n",
            "Epoch 4: val_loss improved from 1.46873 to 1.31378, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.3220 - accuracy: 0.6136 - val_loss: 1.3138 - val_accuracy: 0.6042 - 36s/epoch - 22ms/step\n",
            "Epoch 5/30\n",
            "\n",
            "Epoch 5: val_loss improved from 1.31378 to 1.16500, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.2212 - accuracy: 0.6408 - val_loss: 1.1650 - val_accuracy: 0.6443 - 36s/epoch - 22ms/step\n",
            "Epoch 6/30\n",
            "\n",
            "Epoch 6: val_loss improved from 1.16500 to 1.07883, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.1381 - accuracy: 0.6673 - val_loss: 1.0788 - val_accuracy: 0.6867 - 36s/epoch - 22ms/step\n",
            "Epoch 7/30\n",
            "\n",
            "Epoch 7: val_loss did not improve from 1.07883\n",
            "1632/1632 - 35s - loss: 1.0769 - accuracy: 0.6814 - val_loss: 1.0866 - val_accuracy: 0.6691 - 35s/epoch - 22ms/step\n",
            "Epoch 8/30\n",
            "\n",
            "Epoch 8: val_loss improved from 1.07883 to 1.07321, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 1.0170 - accuracy: 0.7010 - val_loss: 1.0732 - val_accuracy: 0.6858 - 36s/epoch - 22ms/step\n",
            "Epoch 9/30\n",
            "\n",
            "Epoch 9: val_loss improved from 1.07321 to 0.91643, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.9742 - accuracy: 0.7129 - val_loss: 0.9164 - val_accuracy: 0.7443 - 36s/epoch - 22ms/step\n",
            "Epoch 10/30\n",
            "\n",
            "Epoch 10: val_loss did not improve from 0.91643\n",
            "1632/1632 - 35s - loss: 0.9271 - accuracy: 0.7275 - val_loss: 0.9316 - val_accuracy: 0.7290 - 35s/epoch - 22ms/step\n",
            "Epoch 11/30\n",
            "\n",
            "Epoch 11: val_loss did not improve from 0.91643\n",
            "1632/1632 - 35s - loss: 0.9001 - accuracy: 0.7344 - val_loss: 0.9695 - val_accuracy: 0.7071 - 35s/epoch - 22ms/step\n",
            "Epoch 12/30\n",
            "\n",
            "Epoch 12: val_loss improved from 0.91643 to 0.86989, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.8642 - accuracy: 0.7431 - val_loss: 0.8699 - val_accuracy: 0.7443 - 36s/epoch - 22ms/step\n",
            "Epoch 13/30\n",
            "\n",
            "Epoch 13: val_loss improved from 0.86989 to 0.86037, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.8386 - accuracy: 0.7539 - val_loss: 0.8604 - val_accuracy: 0.7482 - 36s/epoch - 22ms/step\n",
            "Epoch 14/30\n",
            "\n",
            "Epoch 14: val_loss improved from 0.86037 to 0.80611, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.8050 - accuracy: 0.7618 - val_loss: 0.8061 - val_accuracy: 0.7668 - 36s/epoch - 22ms/step\n",
            "Epoch 15/30\n",
            "\n",
            "Epoch 15: val_loss did not improve from 0.80611\n",
            "1632/1632 - 35s - loss: 0.7839 - accuracy: 0.7680 - val_loss: 0.8083 - val_accuracy: 0.7594 - 35s/epoch - 22ms/step\n",
            "Epoch 16/30\n",
            "\n",
            "Epoch 16: val_loss improved from 0.80611 to 0.80314, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.7683 - accuracy: 0.7735 - val_loss: 0.8031 - val_accuracy: 0.7561 - 36s/epoch - 22ms/step\n",
            "Epoch 17/30\n",
            "\n",
            "Epoch 17: val_loss did not improve from 0.80314\n",
            "1632/1632 - 35s - loss: 0.7470 - accuracy: 0.7790 - val_loss: 0.8415 - val_accuracy: 0.7296 - 35s/epoch - 22ms/step\n",
            "Epoch 18/30\n",
            "\n",
            "Epoch 18: val_loss improved from 0.80314 to 0.72969, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.7263 - accuracy: 0.7832 - val_loss: 0.7297 - val_accuracy: 0.7856 - 36s/epoch - 22ms/step\n",
            "Epoch 19/30\n",
            "\n",
            "Epoch 19: val_loss improved from 0.72969 to 0.70989, saving model to /content/best_model.h5\n",
            "1632/1632 - 37s - loss: 0.7117 - accuracy: 0.7880 - val_loss: 0.7099 - val_accuracy: 0.7951 - 37s/epoch - 22ms/step\n",
            "Epoch 20/30\n",
            "\n",
            "Epoch 20: val_loss did not improve from 0.70989\n",
            "1632/1632 - 36s - loss: 0.6908 - accuracy: 0.7964 - val_loss: 0.7707 - val_accuracy: 0.7706 - 36s/epoch - 22ms/step\n",
            "Epoch 21/30\n",
            "\n",
            "Epoch 21: val_loss did not improve from 0.70989\n",
            "1632/1632 - 35s - loss: 0.6766 - accuracy: 0.8005 - val_loss: 0.7155 - val_accuracy: 0.7748 - 35s/epoch - 22ms/step\n",
            "Epoch 22/30\n",
            "\n",
            "Epoch 22: val_loss improved from 0.70989 to 0.68283, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.6629 - accuracy: 0.8053 - val_loss: 0.6828 - val_accuracy: 0.7972 - 36s/epoch - 22ms/step\n",
            "Epoch 23/30\n",
            "\n",
            "Epoch 23: val_loss improved from 0.68283 to 0.67354, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.6488 - accuracy: 0.8087 - val_loss: 0.6735 - val_accuracy: 0.8011 - 36s/epoch - 22ms/step\n",
            "Epoch 24/30\n",
            "\n",
            "Epoch 24: val_loss did not improve from 0.67354\n",
            "1632/1632 - 35s - loss: 0.6418 - accuracy: 0.8107 - val_loss: 0.7037 - val_accuracy: 0.7778 - 35s/epoch - 22ms/step\n",
            "Epoch 25/30\n",
            "\n",
            "Epoch 25: val_loss did not improve from 0.67354\n",
            "1632/1632 - 35s - loss: 0.6239 - accuracy: 0.8161 - val_loss: 0.6752 - val_accuracy: 0.7989 - 35s/epoch - 22ms/step\n",
            "Epoch 26/30\n",
            "\n",
            "Epoch 26: val_loss improved from 0.67354 to 0.62337, saving model to /content/best_model.h5\n",
            "1632/1632 - 36s - loss: 0.6118 - accuracy: 0.8189 - val_loss: 0.6234 - val_accuracy: 0.8171 - 36s/epoch - 22ms/step\n",
            "Epoch 27/30\n",
            "\n",
            "Epoch 27: val_loss did not improve from 0.62337\n",
            "1632/1632 - 36s - loss: 0.6041 - accuracy: 0.8218 - val_loss: 0.6237 - val_accuracy: 0.8166 - 36s/epoch - 22ms/step\n",
            "Epoch 28/30\n",
            "\n",
            "Epoch 28: val_loss did not improve from 0.62337\n",
            "1632/1632 - 36s - loss: 0.5920 - accuracy: 0.8234 - val_loss: 0.6537 - val_accuracy: 0.7966 - 36s/epoch - 22ms/step\n",
            "Epoch 29/30\n",
            "\n",
            "Epoch 29: val_loss did not improve from 0.62337\n",
            "1632/1632 - 36s - loss: 0.5822 - accuracy: 0.8271 - val_loss: 0.6438 - val_accuracy: 0.8019 - 36s/epoch - 22ms/step\n",
            "Epoch 30/30\n",
            "\n",
            "Epoch 30: val_loss improved from 0.62337 to 0.61870, saving model to /content/best_model.h5\n",
            "1632/1632 - 37s - loss: 0.5748 - accuracy: 0.8284 - val_loss: 0.6187 - val_accuracy: 0.8056 - 37s/epoch - 22ms/step\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7fb3acd68910>"
            ]
          },
          "metadata": {},
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "resnet_fe3 = [1.8660, 1.5587, 1.4687, 1.3138, 1.1650, 1.0788, 1.0866, 1.0732, 0.9164, 0.9316, 0.9695, 0.8699, 0.8604, 0.8061, 0.8083, 0.8031, 0.8415, 0.7297, 0.7099, 0.7707, 0.7155, 0.6828, 0.6735, 0.7037, 0.6752, 0.6234, 0.6237, 0.6537, 0.6438, 0.6187]"
      ],
      "metadata": {
        "id": "-E0UwaBNye9N"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "score = feature_extraction.evaluate(x = X_test, y = y_cat_test, verbose = 1)\n",
        "print('Accuracy for test images:', round(score[1]*100, 3), '%')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f5624a7d-872e-41f6-c4d7-d0a1aaf8c394",
        "id": "54ocoLRyeMmE"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "544/544 [==============================] - 10s 18ms/step - loss: 0.6125 - accuracy: 0.8067\n",
            "Accuracy for test images: 80.672 %\n"
          ]
        }
      ]
    }
  ]
}