SLR98_trainer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "pycharm": {
     "is_executing": true
    }
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "Created on Thu Nov 26 19:39:00 2020(rewrite)\n",
    "\n",
    "@author: anshtyagi\n",
    "\"\"\"\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "import keras\n",
    "from keras.layers import Conv2D, MaxPool2D, Flatten, Dense, Dropout, BatchNormalization\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras import regularizers\n",
    "from sklearn.model_selection import train_test_split\n",
    "import cv2\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.utils.vis_utils import plot_model\n",
    "from IPython.display import Image\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading data from: 0 ~ A ~ B ~ C ~ D ~ E ~ F ~ G ~ H ~ I ~ J ~ K ~ L ~ M ~ N ~ O ~ P ~ Q ~ R ~ S ~ T ~ U ~ V ~ W ~ X ~ Y ~ Z ~ "
     ]
    }
   ],
   "source": [
    "train_set = \"data/train\"\n",
    "#PreProcessing Module\n",
    "if not os.path.exists(\"preprocess_data\"):\n",
    "    os.makedirs(\"preprocess_data\")\n",
    "if not os.path.exists(\"preprocess_data/train\"):\n",
    "    os.makedirs(\"preprocess_data/train\")\n",
    "minVal = 70\n",
    "def processing(path):\n",
    "    frame = cv2.imread(path)\n",
    "\n",
    "    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)\n",
    "    blur = cv2.GaussianBlur(gray, (5, 5), 2)\n",
    "\n",
    "    th3 = cv2.adaptiveThreshold(blur, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 11, 2)\n",
    "    ret, res = cv2.threshold(th3, minVal, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU)\n",
    "    return res\n",
    "\n",
    "\n",
    "def load_dataset():\n",
    "    IMG_SIZE = 128,128\n",
    "    print(\"Loading data from: \",end=\"\")\n",
    "    for folder in sorted(os.listdir(train_set)):\n",
    "        print(folder,end=\" ~ \")\n",
    "        for image in sorted(os.listdir(train_set+\"/\"+folder)):\n",
    "            temp_image_path = train_set+\"/\"+folder+\"/\"+image\n",
    "            temp_image = processing(temp_image_path)\n",
    "            if not os.path.exists(\"preprocess_data/train/\"+folder):\n",
    "                os.makedirs(\"preprocess_data/train/\"+folder)\n",
    "            final_path = \"preprocess_data/train/\"+folder\n",
    "            cv2.imwrite(os.path.join(final_path,image),temp_image)\n",
    "\n",
    "load_dataset()    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 15585 images belonging to 27 classes.\n",
      "Found 3883 images belonging to 27 classes.\n"
     ]
    }
   ],
   "source": [
    "train_set = \"preprocess_data/train\"\n",
    "IMG_SIZE = 128\n",
    "\n",
    "#train-validation split \n",
    "train_datagen = ImageDataGenerator(rescale=1./255,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,validation_split=0.2)\n",
    "training_set = train_datagen.flow_from_directory(train_set,target_size=(IMG_SIZE,IMG_SIZE),batch_size=64,color_mode=\"grayscale\",class_mode='categorical',subset=\"training\")\n",
    "validation_set = train_datagen.flow_from_directory(train_set,target_size=(IMG_SIZE,IMG_SIZE),subset=\"validation\",batch_size=64,color_mode='grayscale',class_mode='categorical')\n"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Created\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 128, 128, 64)      640       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 64, 64, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 64, 64, 16)        1040      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 21, 21, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 21, 21, 16)        2320      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 10, 10, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 10, 10, 32)        544       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 5, 5, 32)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 5, 5, 64)          51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 2, 2, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 128)               32896     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 64)                8256      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 27)                1755      \n",
      "=================================================================\n",
      "Total params: 98,715\n",
      "Trainable params: 98,715\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/100\n",
      "243/243 [==============================] - 124s 511ms/step - loss: 2.8962 - accuracy: 0.1227 - val_loss: 1.9146 - val_accuracy: 0.3807\n",
      "Epoch 2/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 1.4327 - accuracy: 0.5181 - val_loss: 0.8584 - val_accuracy: 0.7081\n",
      "Epoch 3/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.8673 - accuracy: 0.7039 - val_loss: 0.5407 - val_accuracy: 0.8581\n",
      "Epoch 4/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.6744 - accuracy: 0.7692 - val_loss: 0.4680 - val_accuracy: 0.8695\n",
      "Epoch 5/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.5626 - accuracy: 0.8093 - val_loss: 0.3496 - val_accuracy: 0.9013\n",
      "Epoch 6/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.4741 - accuracy: 0.8394 - val_loss: 0.3241 - val_accuracy: 0.9047\n",
      "Epoch 7/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.4120 - accuracy: 0.8606 - val_loss: 0.2774 - val_accuracy: 0.9203\n",
      "Epoch 8/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.3740 - accuracy: 0.8751 - val_loss: 0.2611 - val_accuracy: 0.9294\n",
      "Epoch 9/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.3282 - accuracy: 0.8922 - val_loss: 0.2229 - val_accuracy: 0.9380\n",
      "Epoch 10/100\n",
      "243/243 [==============================] - 123s 505ms/step - loss: 0.3029 - accuracy: 0.8993 - val_loss: 0.2254 - val_accuracy: 0.9443\n",
      "Epoch 11/100\n",
      "243/243 [==============================] - 122s 501ms/step - loss: 0.2718 - accuracy: 0.9095 - val_loss: 0.2173 - val_accuracy: 0.9438\n",
      "Epoch 12/100\n",
      "243/243 [==============================] - 121s 496ms/step - loss: 0.2559 - accuracy: 0.9183 - val_loss: 0.1823 - val_accuracy: 0.9495\n",
      "Epoch 13/100\n",
      "243/243 [==============================] - 122s 500ms/step - loss: 0.2358 - accuracy: 0.9223 - val_loss: 0.1695 - val_accuracy: 0.9568\n",
      "Epoch 14/100\n",
      "243/243 [==============================] - 122s 502ms/step - loss: 0.2254 - accuracy: 0.9246 - val_loss: 0.2349 - val_accuracy: 0.9339\n",
      "Epoch 15/100\n",
      "243/243 [==============================] - 121s 500ms/step - loss: 0.2141 - accuracy: 0.9302 - val_loss: 0.1792 - val_accuracy: 0.9552\n",
      "Epoch 16/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.1924 - accuracy: 0.9392 - val_loss: 0.1771 - val_accuracy: 0.9583\n",
      "Epoch 17/100\n",
      "243/243 [==============================] - 124s 511ms/step - loss: 0.1880 - accuracy: 0.9400 - val_loss: 0.1755 - val_accuracy: 0.9599\n",
      "Epoch 18/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.1681 - accuracy: 0.9481 - val_loss: 0.1540 - val_accuracy: 0.9620\n",
      "Epoch 19/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.1715 - accuracy: 0.9447 - val_loss: 0.1711 - val_accuracy: 0.9604\n",
      "Epoch 20/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.1652 - accuracy: 0.9459 - val_loss: 0.1496 - val_accuracy: 0.9604\n",
      "Epoch 21/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.1642 - accuracy: 0.9473 - val_loss: 0.1269 - val_accuracy: 0.9685\n",
      "Epoch 22/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.1512 - accuracy: 0.9534 - val_loss: 0.1357 - val_accuracy: 0.9701\n",
      "Epoch 23/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.1515 - accuracy: 0.9543 - val_loss: 0.1830 - val_accuracy: 0.9589\n",
      "Epoch 24/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.1357 - accuracy: 0.9553 - val_loss: 0.1445 - val_accuracy: 0.9633\n",
      "Epoch 25/100\n",
      "243/243 [==============================] - 122s 503ms/step - loss: 0.1391 - accuracy: 0.9576 - val_loss: 0.1068 - val_accuracy: 0.9719\n",
      "Epoch 26/100\n",
      "243/243 [==============================] - 122s 503ms/step - loss: 0.1204 - accuracy: 0.9622 - val_loss: 0.1270 - val_accuracy: 0.9620\n",
      "Epoch 27/100\n",
      "243/243 [==============================] - 122s 503ms/step - loss: 0.1301 - accuracy: 0.9594 - val_loss: 0.1271 - val_accuracy: 0.9695\n",
      "Epoch 28/100\n",
      "243/243 [==============================] - 123s 504ms/step - loss: 0.1210 - accuracy: 0.9603 - val_loss: 0.0959 - val_accuracy: 0.9755\n",
      "Epoch 29/100\n",
      "243/243 [==============================] - 122s 504ms/step - loss: 0.1233 - accuracy: 0.9599 - val_loss: 0.1240 - val_accuracy: 0.9703\n",
      "Epoch 30/100\n",
      "243/243 [==============================] - 123s 505ms/step - loss: 0.1013 - accuracy: 0.9666 - val_loss: 0.1113 - val_accuracy: 0.9727\n",
      "Epoch 31/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.1163 - accuracy: 0.9628 - val_loss: 0.1114 - val_accuracy: 0.9745\n",
      "Epoch 32/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.1166 - accuracy: 0.9639 - val_loss: 0.0812 - val_accuracy: 0.9773\n",
      "Epoch 33/100\n",
      "243/243 [==============================] - 127s 524ms/step - loss: 0.1018 - accuracy: 0.9673 - val_loss: 0.1601 - val_accuracy: 0.9604\n",
      "Epoch 34/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.1053 - accuracy: 0.9674 - val_loss: 0.1166 - val_accuracy: 0.9727\n",
      "Epoch 35/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.1013 - accuracy: 0.9667 - val_loss: 0.0999 - val_accuracy: 0.9737\n",
      "Epoch 36/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.0884 - accuracy: 0.9715 - val_loss: 0.1131 - val_accuracy: 0.9716\n",
      "Epoch 37/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.0898 - accuracy: 0.9714 - val_loss: 0.0881 - val_accuracy: 0.9779\n",
      "Epoch 38/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.0876 - accuracy: 0.9733 - val_loss: 0.0811 - val_accuracy: 0.9781\n",
      "Epoch 39/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "243/243 [==============================] - 124s 509ms/step - loss: 0.0966 - accuracy: 0.9698 - val_loss: 0.0994 - val_accuracy: 0.9742\n",
      "Epoch 40/100\n",
      "243/243 [==============================] - 124s 511ms/step - loss: 0.0940 - accuracy: 0.9705 - val_loss: 0.1359 - val_accuracy: 0.9680\n",
      "Epoch 41/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0926 - accuracy: 0.9718 - val_loss: 0.0995 - val_accuracy: 0.9745\n",
      "Epoch 42/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.0819 - accuracy: 0.9739 - val_loss: 0.1145 - val_accuracy: 0.9716\n",
      "Epoch 43/100\n",
      "243/243 [==============================] - 124s 512ms/step - loss: 0.0846 - accuracy: 0.9733 - val_loss: 0.1033 - val_accuracy: 0.9740\n",
      "Epoch 44/100\n",
      "243/243 [==============================] - 124s 509ms/step - loss: 0.0873 - accuracy: 0.9742 - val_loss: 0.0905 - val_accuracy: 0.9758\n",
      "Epoch 45/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.0748 - accuracy: 0.9773 - val_loss: 0.0972 - val_accuracy: 0.9789\n",
      "Epoch 46/100\n",
      "243/243 [==============================] - 125s 513ms/step - loss: 0.0822 - accuracy: 0.9727 - val_loss: 0.0829 - val_accuracy: 0.9786\n",
      "Epoch 47/100\n",
      "243/243 [==============================] - 124s 512ms/step - loss: 0.0797 - accuracy: 0.9751 - val_loss: 0.1061 - val_accuracy: 0.9740\n",
      "Epoch 48/100\n",
      "243/243 [==============================] - 123s 505ms/step - loss: 0.0707 - accuracy: 0.9780 - val_loss: 0.0960 - val_accuracy: 0.9789\n",
      "Epoch 49/100\n",
      "243/243 [==============================] - 122s 500ms/step - loss: 0.0732 - accuracy: 0.9782 - val_loss: 0.1032 - val_accuracy: 0.9755\n",
      "Epoch 50/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.0847 - accuracy: 0.9747 - val_loss: 0.0933 - val_accuracy: 0.9755\n",
      "Epoch 51/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0704 - accuracy: 0.9782 - val_loss: 0.0928 - val_accuracy: 0.9786\n",
      "Epoch 52/100\n",
      "243/243 [==============================] - 123s 505ms/step - loss: 0.0726 - accuracy: 0.9776 - val_loss: 0.1053 - val_accuracy: 0.9799\n",
      "Epoch 53/100\n",
      "243/243 [==============================] - 123s 506ms/step - loss: 0.0736 - accuracy: 0.9776 - val_loss: 0.1016 - val_accuracy: 0.9724\n",
      "Epoch 54/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.0679 - accuracy: 0.9788 - val_loss: 0.0706 - val_accuracy: 0.9831\n",
      "Epoch 55/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0641 - accuracy: 0.9794 - val_loss: 0.0868 - val_accuracy: 0.9771\n",
      "Epoch 56/100\n",
      "243/243 [==============================] - 123s 505ms/step - loss: 0.0746 - accuracy: 0.9762 - val_loss: 0.1200 - val_accuracy: 0.9690\n",
      "Epoch 57/100\n",
      "243/243 [==============================] - 122s 501ms/step - loss: 0.0658 - accuracy: 0.9794 - val_loss: 0.1038 - val_accuracy: 0.9789\n",
      "Epoch 58/100\n",
      "243/243 [==============================] - 122s 501ms/step - loss: 0.0648 - accuracy: 0.9792 - val_loss: 0.0737 - val_accuracy: 0.9820\n",
      "Epoch 59/100\n",
      "243/243 [==============================] - 124s 511ms/step - loss: 0.0687 - accuracy: 0.9798 - val_loss: 0.0918 - val_accuracy: 0.9779\n",
      "Epoch 60/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.0633 - accuracy: 0.9807 - val_loss: 0.0733 - val_accuracy: 0.9805\n",
      "Epoch 61/100\n",
      "243/243 [==============================] - 124s 510ms/step - loss: 0.0658 - accuracy: 0.9820 - val_loss: 0.0855 - val_accuracy: 0.9812\n",
      "Epoch 62/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0567 - accuracy: 0.9825 - val_loss: 0.0756 - val_accuracy: 0.9815\n",
      "Epoch 63/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0597 - accuracy: 0.9824 - val_loss: 0.0793 - val_accuracy: 0.9786\n",
      "Epoch 64/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.0596 - accuracy: 0.9811 - val_loss: 0.0811 - val_accuracy: 0.9805\n",
      "Epoch 65/100\n",
      "243/243 [==============================] - 124s 508ms/step - loss: 0.0560 - accuracy: 0.9833 - val_loss: 0.0919 - val_accuracy: 0.9766\n",
      "Epoch 66/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.0639 - accuracy: 0.9813 - val_loss: 0.0787 - val_accuracy: 0.9823\n",
      "Epoch 67/100\n",
      "243/243 [==============================] - 123s 508ms/step - loss: 0.0559 - accuracy: 0.9817 - val_loss: 0.0741 - val_accuracy: 0.9828\n",
      "Epoch 68/100\n",
      "243/243 [==============================] - 125s 514ms/step - loss: 0.0635 - accuracy: 0.9811 - val_loss: 0.0891 - val_accuracy: 0.9763\n",
      "Epoch 69/100\n",
      "243/243 [==============================] - 127s 523ms/step - loss: 0.0554 - accuracy: 0.9833 - val_loss: 0.1161 - val_accuracy: 0.9719\n",
      "Epoch 70/100\n",
      "243/243 [==============================] - 126s 518ms/step - loss: 0.0623 - accuracy: 0.9796 - val_loss: 0.0862 - val_accuracy: 0.9810\n",
      "Epoch 71/100\n",
      "243/243 [==============================] - 122s 501ms/step - loss: 0.0568 - accuracy: 0.9816 - val_loss: 0.0813 - val_accuracy: 0.9810\n",
      "Epoch 72/100\n",
      "243/243 [==============================] - 123s 507ms/step - loss: 0.0498 - accuracy: 0.9842 - val_loss: 0.0842 - val_accuracy: 0.9802\n",
      "Epoch 73/100\n",
      "243/243 [==============================] - 125s 515ms/step - loss: 0.0571 - accuracy: 0.9833 - val_loss: 0.0697 - val_accuracy: 0.9839\n",
      "Epoch 74/100\n",
      "243/243 [==============================] - 130s 534ms/step - loss: 0.0580 - accuracy: 0.9813 - val_loss: 0.0803 - val_accuracy: 0.9789\n",
      "Epoch 75/100\n",
      "243/243 [==============================] - 130s 533ms/step - loss: 0.0543 - accuracy: 0.9828 - val_loss: 0.0882 - val_accuracy: 0.9776\n",
      "Epoch 76/100\n",
      "243/243 [==============================] - 128s 529ms/step - loss: 0.0533 - accuracy: 0.9837 - val_loss: 0.0956 - val_accuracy: 0.9755\n",
      "Epoch 77/100\n",
      "243/243 [==============================] - 131s 538ms/step - loss: 0.0520 - accuracy: 0.9840 - val_loss: 0.0651 - val_accuracy: 0.9828\n",
      "Epoch 78/100\n",
      "243/243 [==============================] - 130s 535ms/step - loss: 0.0532 - accuracy: 0.9843 - val_loss: 0.0901 - val_accuracy: 0.9818\n",
      "Epoch 79/100\n",
      "243/243 [==============================] - 131s 540ms/step - loss: 0.0523 - accuracy: 0.9840 - val_loss: 0.0717 - val_accuracy: 0.9810\n",
      "Epoch 80/100\n",
      "243/243 [==============================] - 130s 533ms/step - loss: 0.0523 - accuracy: 0.9840 - val_loss: 0.0663 - val_accuracy: 0.9820\n",
      "Epoch 81/100\n",
      "243/243 [==============================] - 129s 531ms/step - loss: 0.0469 - accuracy: 0.9852 - val_loss: 0.0966 - val_accuracy: 0.9729\n",
      "Epoch 82/100\n",
      "243/243 [==============================] - 127s 522ms/step - loss: 0.0533 - accuracy: 0.9827 - val_loss: 0.0806 - val_accuracy: 0.9802\n",
      "Epoch 83/100\n",
      "243/243 [==============================] - 131s 539ms/step - loss: 0.0536 - accuracy: 0.9838 - val_loss: 0.0822 - val_accuracy: 0.9786\n",
      "Epoch 84/100\n",
      "243/243 [==============================] - 136s 561ms/step - loss: 0.0463 - accuracy: 0.9861 - val_loss: 0.0720 - val_accuracy: 0.9841\n",
      "Epoch 85/100\n",
      "243/243 [==============================] - 137s 563ms/step - loss: 0.0425 - accuracy: 0.9863 - val_loss: 0.0791 - val_accuracy: 0.9794\n",
      "Epoch 86/100\n",
      "243/243 [==============================] - 139s 574ms/step - loss: 0.0452 - accuracy: 0.9864 - val_loss: 0.0757 - val_accuracy: 0.9779\n",
      "Epoch 87/100\n",
      "243/243 [==============================] - 145s 597ms/step - loss: 0.0487 - accuracy: 0.9853 - val_loss: 0.0634 - val_accuracy: 0.9805\n",
      "Epoch 88/100\n",
      "243/243 [==============================] - 144s 594ms/step - loss: 0.0516 - accuracy: 0.9845 - val_loss: 0.0655 - val_accuracy: 0.9812\n",
      "Epoch 89/100\n",
      "243/243 [==============================] - 144s 595ms/step - loss: 0.0443 - accuracy: 0.9861 - val_loss: 0.0541 - val_accuracy: 0.9865\n",
      "Epoch 90/100\n",
      "243/243 [==============================] - 142s 584ms/step - loss: 0.0498 - accuracy: 0.9841 - val_loss: 0.1051 - val_accuracy: 0.9776\n",
      "Epoch 91/100\n",
      "243/243 [==============================] - 144s 592ms/step - loss: 0.0456 - accuracy: 0.9851 - val_loss: 0.0753 - val_accuracy: 0.9831\n",
      "Epoch 92/100\n",
      "243/243 [==============================] - 139s 571ms/step - loss: 0.0454 - accuracy: 0.9858 - val_loss: 0.0634 - val_accuracy: 0.9807\n",
      "Epoch 93/100\n",
      "243/243 [==============================] - 140s 576ms/step - loss: 0.0393 - accuracy: 0.9879 - val_loss: 0.0827 - val_accuracy: 0.9794\n",
      "Epoch 94/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "243/243 [==============================] - 138s 567ms/step - loss: 0.0513 - accuracy: 0.9838 - val_loss: 0.0592 - val_accuracy: 0.9854\n",
      "Epoch 95/100\n",
      "243/243 [==============================] - 150s 618ms/step - loss: 0.0395 - accuracy: 0.9878 - val_loss: 0.0695 - val_accuracy: 0.9870\n",
      "Epoch 96/100\n",
      "243/243 [==============================] - 145s 598ms/step - loss: 0.0432 - accuracy: 0.9861 - val_loss: 0.0570 - val_accuracy: 0.9836\n",
      "Epoch 97/100\n",
      "243/243 [==============================] - 148s 610ms/step - loss: 0.0503 - accuracy: 0.9855 - val_loss: 0.0441 - val_accuracy: 0.9867\n",
      "Epoch 98/100\n",
      "243/243 [==============================] - 190s 782ms/step - loss: 0.0443 - accuracy: 0.9870 - val_loss: 0.0765 - val_accuracy: 0.9831\n",
      "Epoch 99/100\n",
      "243/243 [==============================] - 184s 757ms/step - loss: 0.0431 - accuracy: 0.9867 - val_loss: 0.0658 - val_accuracy: 0.9849\n",
      "Epoch 100/100\n",
      "243/243 [==============================] - 184s 758ms/step - loss: 0.0437 - accuracy: 0.9869 - val_loss: 0.0729 - val_accuracy: 0.9810\n"
     ]
    }
   ],
   "source": [
    "def create_model():\n",
    "    model = Sequential()\n",
    "\n",
    "    #First Layer of Conv2D\n",
    "    model.add(Conv2D(64,kernel_size=[3,3],padding=\"same\",activation=\"relu\",input_shape=(128,128,1)))\n",
    "    model.add(MaxPool2D(pool_size=[2,2]))\n",
    "\n",
    "    #Second Layer\n",
    "    model.add(Conv2D(16,kernel_size=[1,1],padding=\"same\",activation=\"relu\"))\n",
    "    model.add(MaxPool2D(pool_size=[3,3]))\n",
    "\n",
    "    #Third Layer\n",
    "    model.add(Conv2D(16,kernel_size=[3,3],padding=\"same\",activation=\"relu\"))\n",
    "    model.add(MaxPool2D(pool_size=[2,2]))\n",
    "\n",
    "    #Fourth Layer\n",
    "    model.add(Conv2D(32,kernel_size=[1,1],padding=\"same\",activation=\"relu\"))\n",
    "    model.add(MaxPool2D(pool_size=[2,2]))\n",
    "\n",
    "    #Fifth Layer\n",
    "    model.add(Conv2D(64,kernel_size=[5,5],padding=\"same\",activation=\"relu\"))\n",
    "    model.add(MaxPool2D(pool_size=[2,2]))\n",
    "\n",
    "    #Flatten\n",
    "    model.add(Flatten())\n",
    "\n",
    "    #Dropout\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(128,activation=\"relu\"))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(64,activation=\"relu\"))\n",
    "    model.add(Dense(27,activation=\"softmax\"))\n",
    "\n",
    "    model.compile(optimizer='adam', loss=keras.losses.categorical_crossentropy, metrics=[\"accuracy\"])\n",
    "\n",
    "    print(\"Model Created\")\n",
    "    model.summary()\n",
    "    model_hist = model.fit(training_set,steps_per_epoch=15585//64,epochs=100,validation_data=validation_set,validation_steps=3883//64)\n",
    "    return model,model_hist\n",
    "\n",
    "\n",
    "model,model_hist = create_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-5-dc452f020b4c>:1: Model.evaluate_generator (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use Model.evaluate, which supports generators.\n",
      "WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 3883 batches). You may need to use the repeat() function when building your dataset.\n",
      "\n",
      "Evaluation Accuracy =  98.12% \n",
      "Evaluation loss =  0.070325\n"
     ]
    }
   ],
   "source": [
    "evaluate_metrics = model.evaluate_generator(validation_set,3883)\n",
    "print(\"\\nEvaluation Accuracy = \", \"{:.2f}%\".format(evaluate_metrics[1]*100),\"\\nEvaluation loss = \" ,\"{:.6f}\".format(evaluate_metrics[0]))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Accuracy Plot\n",
    "plt.plot(model_hist.history[\"accuracy\"])\n",
    "plt.plot(model_hist.history['val_accuracy'])\n",
    "plt.legend(['train', 'test'], loc='lower right')\n",
    "plt.title('accuracy plot - train vs test')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('accuracy')\n",
    "plt.show()\n",
    "\n",
    "#Loss Plot\n",
    "plt.plot(model_hist.history['loss'])\n",
    "plt.plot(model_hist.history['val_loss'])\n",
    "plt.legend(['training loss', 'validation loss'], loc = 'upper right')\n",
    "plt.title('loss plot - training vs vaidation')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model Saved as json\n",
      "Weights saved\n",
      "Model Saved as h5\n"
     ]
    }
   ],
   "source": [
    "model_json = model.to_json()\n",
    "with open(\"SLR98-100.json\", \"w\") as json_file:\n",
    "    json_file.write(model_json)\n",
    "print('Model Saved as json')\n",
    "model.save_weights('SLR98_100-wt.h5')\n",
    "print('Weights saved')\n",
    "model.save(\"SLR98_100.h5\")\n",
    "print(\"Model Saved as h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "PyCharm (SLR98)",
   "language": "python",
   "name": "pycharm-f123a4d3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}