Transfer Learning for Training.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "242350f6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries\n",
    "\n",
    "from tensorflow.keras.layers import Input, Lambda, Dense, Flatten\n",
    "from tensorflow.keras.models import Sequential, Model,load_model\n",
    "from tensorflow.keras.applications.resnet import ResNet50,preprocess_input\n",
    "from tensorflow.keras.applications.vgg16 import VGG16\n",
    "from tensorflow.keras.applications.inception_v3 import InceptionV3\n",
    "from tensorflow.keras.applications.xception import Xception\n",
    "from tensorflow.keras.preprocessing import image\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator,load_img\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "import os\n",
    "import numpy\n",
    "from glob import glob\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "99c17fb1",
   "metadata": {},
   "outputs": [],
   "source": [
    "IMAGE_SIZE = [224,224]\n",
    "BATCH_SIZE = 32\n",
    "EPOCHS = 50\n",
    "\n",
    "train_path = \"Datasets/Train/\"\n",
    "test_path = \"Datasets/Test/\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "e6558a5e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_model(model_name, best_model_name):\n",
    "    \n",
    "    base_model = model_name(input_shape=IMAGE_SIZE + [3],weights='imagenet', include_top=False)\n",
    "    \n",
    "    # don't train existing weights\n",
    "    for layer in base_model.layers:\n",
    "        layer.trainable = False\n",
    "        \n",
    "    # for getting number of output classes\n",
    "    folders = os.listdir(train_path)\n",
    "    \n",
    "    # our layers\n",
    "    x = Flatten()(base_model.output)\n",
    "    \n",
    "    prediction = Dense(len(folders), activation=\"softmax\")(x)\n",
    "    \n",
    "    model = Model(inputs=base_model.input, outputs=prediction)\n",
    "    \n",
    "    model.compile(\n",
    "        loss=\"categorical_crossentropy\",\n",
    "        optimizer=\"adam\",\n",
    "        metrics=[\"accuracy\"]\n",
    "    )\n",
    "    \n",
    "    \n",
    "    train_datagen = ImageDataGenerator(rescale = 1./255,\n",
    "                              shear_range = 0.2,\n",
    "                              zoom_range = 0.2,\n",
    "                              horizontal_flip = True)\n",
    "    \n",
    "    test_datagen = ImageDataGenerator(rescale = 1./255)\n",
    "    \n",
    "    train_set = train_datagen.flow_from_directory(train_path,\n",
    "                                                target_size = IMAGE_SIZE,\n",
    "                                                batch_size = BATCH_SIZE,\n",
    "                                                class_mode = \"categorical\")\n",
    "    \n",
    "    test_set = test_datagen.flow_from_directory(test_path,\n",
    "                                              target_size = IMAGE_SIZE,\n",
    "                                              batch_size = BATCH_SIZE,\n",
    "                                              class_mode = \"categorical\")\n",
    "    \n",
    "    \n",
    "    es = EarlyStopping(monitor='val_loss', mode='min',verbose=1,patience=50)\n",
    "    \n",
    "    mc = ModelCheckpoint(best_model_name, monitor='val_accuracy',mode='max',verbose=1)\n",
    "    \n",
    "    r = model.fit(train_set,\n",
    "              validation_data = test_set,\n",
    "              epochs = EPOCHS,\n",
    "              steps_per_epoch = len(train_set),\n",
    "              validation_steps = len(test_set),\n",
    "              callbacks=[es, mc])\n",
    "    \n",
    "    return r, model, test_set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d495e747",
   "metadata": {},
   "source": [
    "#### I've used four popular Transfer Learning Models - ResNet50, VGG16, InceptionV3 and Xception"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f324ab82",
   "metadata": {},
   "source": [
    "#### ResNet50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "951e6cc7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8006 - accuracy: 0.7000\n",
      "Epoch 00001: saving model to best_resnet_model.h5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\vidhy\\anaconda3\\lib\\site-packages\\keras\\engine\\functional.py:1410: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
      "  layer_config = serialize_layer_fn(layer)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10/10 [==============================] - 8s 809ms/step - loss: 0.8006 - accuracy: 0.7000 - val_loss: 2.3179 - val_accuracy: 0.3100\n",
      "Epoch 2/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8235 - accuracy: 0.6733\n",
      "Epoch 00002: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 7s 682ms/step - loss: 0.8235 - accuracy: 0.6733 - val_loss: 2.1352 - val_accuracy: 0.3100\n",
      "Epoch 3/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.0112 - accuracy: 0.5933\n",
      "Epoch 00003: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 625ms/step - loss: 1.0112 - accuracy: 0.5933 - val_loss: 2.4202 - val_accuracy: 0.3100\n",
      "Epoch 4/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.2025 - accuracy: 0.5800\n",
      "Epoch 00004: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 597ms/step - loss: 1.2025 - accuracy: 0.5800 - val_loss: 2.3509 - val_accuracy: 0.3200\n",
      "Epoch 5/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8321 - accuracy: 0.6733\n",
      "Epoch 00005: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 555ms/step - loss: 0.8321 - accuracy: 0.6733 - val_loss: 2.0376 - val_accuracy: 0.3300\n",
      "Epoch 6/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8339 - accuracy: 0.6733\n",
      "Epoch 00006: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 588ms/step - loss: 0.8339 - accuracy: 0.6733 - val_loss: 2.1216 - val_accuracy: 0.3000\n",
      "Epoch 7/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7999 - accuracy: 0.6867\n",
      "Epoch 00007: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 638ms/step - loss: 0.7999 - accuracy: 0.6867 - val_loss: 1.9826 - val_accuracy: 0.4000\n",
      "Epoch 8/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8557 - accuracy: 0.7033\n",
      "Epoch 00008: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 8s 754ms/step - loss: 0.8557 - accuracy: 0.7033 - val_loss: 2.0781 - val_accuracy: 0.3000\n",
      "Epoch 9/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7427 - accuracy: 0.6867\n",
      "Epoch 00009: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 622ms/step - loss: 0.7427 - accuracy: 0.6867 - val_loss: 1.8257 - val_accuracy: 0.3800\n",
      "Epoch 10/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7993 - accuracy: 0.6667\n",
      "Epoch 00010: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 617ms/step - loss: 0.7993 - accuracy: 0.6667 - val_loss: 1.7460 - val_accuracy: 0.4200\n",
      "Epoch 11/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6655 - accuracy: 0.7600\n",
      "Epoch 00011: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 603ms/step - loss: 0.6655 - accuracy: 0.7600 - val_loss: 1.7197 - val_accuracy: 0.4000\n",
      "Epoch 12/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6947 - accuracy: 0.7367\n",
      "Epoch 00012: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 584ms/step - loss: 0.6947 - accuracy: 0.7367 - val_loss: 1.7083 - val_accuracy: 0.4200\n",
      "Epoch 13/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7987 - accuracy: 0.7300\n",
      "Epoch 00013: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 569ms/step - loss: 0.7987 - accuracy: 0.7300 - val_loss: 2.2937 - val_accuracy: 0.3000\n",
      "Epoch 14/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8093 - accuracy: 0.6633\n",
      "Epoch 00014: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 634ms/step - loss: 0.8093 - accuracy: 0.6633 - val_loss: 1.8433 - val_accuracy: 0.3800\n",
      "Epoch 15/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6673 - accuracy: 0.7467\n",
      "Epoch 00015: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 595ms/step - loss: 0.6673 - accuracy: 0.7467 - val_loss: 1.7090 - val_accuracy: 0.3600\n",
      "Epoch 16/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6483 - accuracy: 0.7500\n",
      "Epoch 00016: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 7s 673ms/step - loss: 0.6483 - accuracy: 0.7500 - val_loss: 1.7407 - val_accuracy: 0.3500\n",
      "Epoch 17/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7222 - accuracy: 0.7167\n",
      "Epoch 00017: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 547ms/step - loss: 0.7222 - accuracy: 0.7167 - val_loss: 1.9217 - val_accuracy: 0.4200\n",
      "Epoch 18/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8582 - accuracy: 0.6700\n",
      "Epoch 00018: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 576ms/step - loss: 0.8582 - accuracy: 0.6700 - val_loss: 2.4839 - val_accuracy: 0.2700\n",
      "Epoch 19/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.9211 - accuracy: 0.6600\n",
      "Epoch 00019: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 532ms/step - loss: 0.9211 - accuracy: 0.6600 - val_loss: 2.0792 - val_accuracy: 0.3600\n",
      "Epoch 20/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.0395 - accuracy: 0.5867\n",
      "Epoch 00020: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 569ms/step - loss: 1.0395 - accuracy: 0.5867 - val_loss: 1.7725 - val_accuracy: 0.4000\n",
      "Epoch 21/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7771 - accuracy: 0.6833\n",
      "Epoch 00021: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 570ms/step - loss: 0.7771 - accuracy: 0.6833 - val_loss: 2.3277 - val_accuracy: 0.3400\n",
      "Epoch 22/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7692 - accuracy: 0.7100\n",
      "Epoch 00022: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 529ms/step - loss: 0.7692 - accuracy: 0.7100 - val_loss: 1.8089 - val_accuracy: 0.4200\n",
      "Epoch 23/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7206 - accuracy: 0.7300\n",
      "Epoch 00023: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 607ms/step - loss: 0.7206 - accuracy: 0.7300 - val_loss: 1.8342 - val_accuracy: 0.4200\n",
      "Epoch 24/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8642 - accuracy: 0.7000\n",
      "Epoch 00024: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 550ms/step - loss: 0.8642 - accuracy: 0.7000 - val_loss: 2.3076 - val_accuracy: 0.3700\n",
      "Epoch 25/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.9418 - accuracy: 0.6500\n",
      "Epoch 00025: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 544ms/step - loss: 0.9418 - accuracy: 0.6500 - val_loss: 2.8947 - val_accuracy: 0.2700\n",
      "Epoch 26/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.1408 - accuracy: 0.5767\n",
      "Epoch 00026: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 589ms/step - loss: 1.1408 - accuracy: 0.5767 - val_loss: 2.8014 - val_accuracy: 0.3000\n",
      "Epoch 27/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.0887 - accuracy: 0.6233\n",
      "Epoch 00027: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 505ms/step - loss: 1.0887 - accuracy: 0.6233 - val_loss: 2.2756 - val_accuracy: 0.4100\n",
      "Epoch 28/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7277 - accuracy: 0.7567\n",
      "Epoch 00028: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 546ms/step - loss: 0.7277 - accuracy: 0.7567 - val_loss: 2.7489 - val_accuracy: 0.3600\n",
      "Epoch 29/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8599 - accuracy: 0.6667\n",
      "Epoch 00029: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 543ms/step - loss: 0.8599 - accuracy: 0.6667 - val_loss: 1.9872 - val_accuracy: 0.3500\n",
      "Epoch 30/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6875 - accuracy: 0.7333\n",
      "Epoch 00030: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 546ms/step - loss: 0.6875 - accuracy: 0.7333 - val_loss: 2.1311 - val_accuracy: 0.4000\n",
      "Epoch 31/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6382 - accuracy: 0.7433\n",
      "Epoch 00031: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 589ms/step - loss: 0.6382 - accuracy: 0.7433 - val_loss: 1.8649 - val_accuracy: 0.4100\n",
      "Epoch 32/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6466 - accuracy: 0.7500\n",
      "Epoch 00032: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 520ms/step - loss: 0.6466 - accuracy: 0.7500 - val_loss: 2.0258 - val_accuracy: 0.3500\n",
      "Epoch 33/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6094 - accuracy: 0.7800\n",
      "Epoch 00033: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 537ms/step - loss: 0.6094 - accuracy: 0.7800 - val_loss: 1.8686 - val_accuracy: 0.3900\n",
      "Epoch 34/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7412 - accuracy: 0.7200\n",
      "Epoch 00034: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 591ms/step - loss: 0.7412 - accuracy: 0.7200 - val_loss: 2.1322 - val_accuracy: 0.3500\n",
      "Epoch 35/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.5372 - accuracy: 0.8167\n",
      "Epoch 00035: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 616ms/step - loss: 0.5372 - accuracy: 0.8167 - val_loss: 2.0763 - val_accuracy: 0.3200\n",
      "Epoch 36/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7024 - accuracy: 0.7233\n",
      "Epoch 00036: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 587ms/step - loss: 0.7024 - accuracy: 0.7233 - val_loss: 1.8789 - val_accuracy: 0.4200\n",
      "Epoch 37/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6227 - accuracy: 0.7633\n",
      "Epoch 00037: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 586ms/step - loss: 0.6227 - accuracy: 0.7633 - val_loss: 2.7798 - val_accuracy: 0.2700\n",
      "Epoch 38/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.0398 - accuracy: 0.6600\n",
      "Epoch 00038: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 563ms/step - loss: 1.0398 - accuracy: 0.6600 - val_loss: 1.8212 - val_accuracy: 0.4200\n",
      "Epoch 39/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7750 - accuracy: 0.7267\n",
      "Epoch 00039: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 566ms/step - loss: 0.7750 - accuracy: 0.7267 - val_loss: 2.3172 - val_accuracy: 0.3500\n",
      "Epoch 40/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6637 - accuracy: 0.7300\n",
      "Epoch 00040: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 590ms/step - loss: 0.6637 - accuracy: 0.7300 - val_loss: 2.5113 - val_accuracy: 0.3700\n",
      "Epoch 41/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8169 - accuracy: 0.7200\n",
      "Epoch 00041: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 555ms/step - loss: 0.8169 - accuracy: 0.7200 - val_loss: 1.8839 - val_accuracy: 0.3900\n",
      "Epoch 42/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6586 - accuracy: 0.7767\n",
      "Epoch 00042: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 577ms/step - loss: 0.6586 - accuracy: 0.7767 - val_loss: 2.1860 - val_accuracy: 0.3700\n",
      "Epoch 43/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.0581 - accuracy: 0.6567\n",
      "Epoch 00043: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 581ms/step - loss: 1.0581 - accuracy: 0.6567 - val_loss: 2.2709 - val_accuracy: 0.3900\n",
      "Epoch 44/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6913 - accuracy: 0.7467\n",
      "Epoch 00044: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 573ms/step - loss: 0.6913 - accuracy: 0.7467 - val_loss: 1.9875 - val_accuracy: 0.4100\n",
      "Epoch 45/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6858 - accuracy: 0.7300\n",
      "Epoch 00045: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 5s 544ms/step - loss: 0.6858 - accuracy: 0.7300 - val_loss: 2.2517 - val_accuracy: 0.3700\n",
      "Epoch 46/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6134 - accuracy: 0.7767\n",
      "Epoch 00046: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 575ms/step - loss: 0.6134 - accuracy: 0.7767 - val_loss: 1.9399 - val_accuracy: 0.4100\n",
      "Epoch 47/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6022 - accuracy: 0.7700\n",
      "Epoch 00047: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 578ms/step - loss: 0.6022 - accuracy: 0.7700 - val_loss: 2.1699 - val_accuracy: 0.3800\n",
      "Epoch 48/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6004 - accuracy: 0.7767\n",
      "Epoch 00048: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 590ms/step - loss: 0.6004 - accuracy: 0.7767 - val_loss: 2.0784 - val_accuracy: 0.4100\n",
      "Epoch 49/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6715 - accuracy: 0.7833\n",
      "Epoch 00049: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 619ms/step - loss: 0.6715 - accuracy: 0.7833 - val_loss: 1.9308 - val_accuracy: 0.4300\n",
      "Epoch 50/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.5596 - accuracy: 0.7933\n",
      "Epoch 00050: saving model to best_resnet_model.h5\n",
      "10/10 [==============================] - 6s 579ms/step - loss: 0.5596 - accuracy: 0.7933 - val_loss: 1.9222 - val_accuracy: 0.3800\n"
     ]
    }
   ],
   "source": [
    "r, model, test_set = create_model(ResNet50, 'best_resnet_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e3c76f33",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_3 (InputLayer)           [(None, 224, 224, 3  0           []                               \n",
      "                                )]                                                                \n",
      "                                                                                                  \n",
      " block1_conv1 (Conv2D)          (None, 111, 111, 32  864         ['input_3[0][0]']                \n",
      "                                )                                                                 \n",
      "                                                                                                  \n",
      " block1_conv1_bn (BatchNormaliz  (None, 111, 111, 32  128        ['block1_conv1[0][0]']           \n",
      " ation)                         )                                                                 \n",
      "                                                                                                  \n",
      " block1_conv1_act (Activation)  (None, 111, 111, 32  0           ['block1_conv1_bn[0][0]']        \n",
      "                                )                                                                 \n",
      "                                                                                                  \n",
      " block1_conv2 (Conv2D)          (None, 109, 109, 64  18432       ['block1_conv1_act[0][0]']       \n",
      "                                )                                                                 \n",
      "                                                                                                  \n",
      " block1_conv2_bn (BatchNormaliz  (None, 109, 109, 64  256        ['block1_conv2[0][0]']           \n",
      " ation)                         )                                                                 \n",
      "                                                                                                  \n",
      " block1_conv2_act (Activation)  (None, 109, 109, 64  0           ['block1_conv2_bn[0][0]']        \n",
      "                                )                                                                 \n",
      "                                                                                                  \n",
      " block2_sepconv1 (SeparableConv  (None, 109, 109, 12  8768       ['block1_conv2_act[0][0]']       \n",
      " 2D)                            8)                                                                \n",
      "                                                                                                  \n",
      " block2_sepconv1_bn (BatchNorma  (None, 109, 109, 12  512        ['block2_sepconv1[0][0]']        \n",
      " lization)                      8)                                                                \n",
      "                                                                                                  \n",
      " block2_sepconv2_act (Activatio  (None, 109, 109, 12  0          ['block2_sepconv1_bn[0][0]']     \n",
      " n)                             8)                                                                \n",
      "                                                                                                  \n",
      " block2_sepconv2 (SeparableConv  (None, 109, 109, 12  17536      ['block2_sepconv2_act[0][0]']    \n",
      " 2D)                            8)                                                                \n",
      "                                                                                                  \n",
      " block2_sepconv2_bn (BatchNorma  (None, 109, 109, 12  512        ['block2_sepconv2[0][0]']        \n",
      " lization)                      8)                                                                \n",
      "                                                                                                  \n",
      " conv2d_8 (Conv2D)              (None, 55, 55, 128)  8192        ['block1_conv2_act[0][0]']       \n",
      "                                                                                                  \n",
      " block2_pool (MaxPooling2D)     (None, 55, 55, 128)  0           ['block2_sepconv2_bn[0][0]']     \n",
      "                                                                                                  \n",
      " batch_normalization_8 (BatchNo  (None, 55, 55, 128)  512        ['conv2d_8[0][0]']               \n",
      " rmalization)                                                                                     \n",
      "                                                                                                  \n",
      " add_24 (Add)                   (None, 55, 55, 128)  0           ['block2_pool[0][0]',            \n",
      "                                                                  'batch_normalization_8[0][0]']  \n",
      "                                                                                                  \n",
      " block3_sepconv1_act (Activatio  (None, 55, 55, 128)  0          ['add_24[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block3_sepconv1 (SeparableConv  (None, 55, 55, 256)  33920      ['block3_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block3_sepconv1_bn (BatchNorma  (None, 55, 55, 256)  1024       ['block3_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block3_sepconv2_act (Activatio  (None, 55, 55, 256)  0          ['block3_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block3_sepconv2 (SeparableConv  (None, 55, 55, 256)  67840      ['block3_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block3_sepconv2_bn (BatchNorma  (None, 55, 55, 256)  1024       ['block3_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " conv2d_9 (Conv2D)              (None, 28, 28, 256)  32768       ['add_24[0][0]']                 \n",
      "                                                                                                  \n",
      " block3_pool (MaxPooling2D)     (None, 28, 28, 256)  0           ['block3_sepconv2_bn[0][0]']     \n",
      "                                                                                                  \n",
      " batch_normalization_9 (BatchNo  (None, 28, 28, 256)  1024       ['conv2d_9[0][0]']               \n",
      " rmalization)                                                                                     \n",
      "                                                                                                  \n",
      " add_25 (Add)                   (None, 28, 28, 256)  0           ['block3_pool[0][0]',            \n",
      "                                                                  'batch_normalization_9[0][0]']  \n",
      "                                                                                                  \n",
      " block4_sepconv1_act (Activatio  (None, 28, 28, 256)  0          ['add_25[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block4_sepconv1 (SeparableConv  (None, 28, 28, 728)  188672     ['block4_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block4_sepconv1_bn (BatchNorma  (None, 28, 28, 728)  2912       ['block4_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block4_sepconv2_act (Activatio  (None, 28, 28, 728)  0          ['block4_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block4_sepconv2 (SeparableConv  (None, 28, 28, 728)  536536     ['block4_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block4_sepconv2_bn (BatchNorma  (None, 28, 28, 728)  2912       ['block4_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " conv2d_10 (Conv2D)             (None, 14, 14, 728)  186368      ['add_25[0][0]']                 \n",
      "                                                                                                  \n",
      " block4_pool (MaxPooling2D)     (None, 14, 14, 728)  0           ['block4_sepconv2_bn[0][0]']     \n",
      "                                                                                                  \n",
      " batch_normalization_10 (BatchN  (None, 14, 14, 728)  2912       ['conv2d_10[0][0]']              \n",
      " ormalization)                                                                                    \n",
      "                                                                                                  \n",
      " add_26 (Add)                   (None, 14, 14, 728)  0           ['block4_pool[0][0]',            \n",
      "                                                                  'batch_normalization_10[0][0]'] \n",
      "                                                                                                  \n",
      " block5_sepconv1_act (Activatio  (None, 14, 14, 728)  0          ['add_26[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block5_sepconv1 (SeparableConv  (None, 14, 14, 728)  536536     ['block5_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block5_sepconv1_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block5_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block5_sepconv2_act (Activatio  (None, 14, 14, 728)  0          ['block5_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block5_sepconv2 (SeparableConv  (None, 14, 14, 728)  536536     ['block5_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block5_sepconv2_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block5_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block5_sepconv3_act (Activatio  (None, 14, 14, 728)  0          ['block5_sepconv2_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block5_sepconv3 (SeparableConv  (None, 14, 14, 728)  536536     ['block5_sepconv3_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block5_sepconv3_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block5_sepconv3[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " add_27 (Add)                   (None, 14, 14, 728)  0           ['block5_sepconv3_bn[0][0]',     \n",
      "                                                                  'add_26[0][0]']                 \n",
      "                                                                                                  \n",
      " block6_sepconv1_act (Activatio  (None, 14, 14, 728)  0          ['add_27[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block6_sepconv1 (SeparableConv  (None, 14, 14, 728)  536536     ['block6_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block6_sepconv1_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block6_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block6_sepconv2_act (Activatio  (None, 14, 14, 728)  0          ['block6_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block6_sepconv2 (SeparableConv  (None, 14, 14, 728)  536536     ['block6_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block6_sepconv2_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block6_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block6_sepconv3_act (Activatio  (None, 14, 14, 728)  0          ['block6_sepconv2_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block6_sepconv3 (SeparableConv  (None, 14, 14, 728)  536536     ['block6_sepconv3_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block6_sepconv3_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block6_sepconv3[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " add_28 (Add)                   (None, 14, 14, 728)  0           ['block6_sepconv3_bn[0][0]',     \n",
      "                                                                  'add_27[0][0]']                 \n",
      "                                                                                                  \n",
      " block7_sepconv1_act (Activatio  (None, 14, 14, 728)  0          ['add_28[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block7_sepconv1 (SeparableConv  (None, 14, 14, 728)  536536     ['block7_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block7_sepconv1_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block7_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block7_sepconv2_act (Activatio  (None, 14, 14, 728)  0          ['block7_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block7_sepconv2 (SeparableConv  (None, 14, 14, 728)  536536     ['block7_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block7_sepconv2_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block7_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block7_sepconv3_act (Activatio  (None, 14, 14, 728)  0          ['block7_sepconv2_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block7_sepconv3 (SeparableConv  (None, 14, 14, 728)  536536     ['block7_sepconv3_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block7_sepconv3_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block7_sepconv3[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " add_29 (Add)                   (None, 14, 14, 728)  0           ['block7_sepconv3_bn[0][0]',     \n",
      "                                                                  'add_28[0][0]']                 \n",
      "                                                                                                  \n",
      " block8_sepconv1_act (Activatio  (None, 14, 14, 728)  0          ['add_29[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block8_sepconv1 (SeparableConv  (None, 14, 14, 728)  536536     ['block8_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block8_sepconv1_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block8_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block8_sepconv2_act (Activatio  (None, 14, 14, 728)  0          ['block8_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block8_sepconv2 (SeparableConv  (None, 14, 14, 728)  536536     ['block8_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block8_sepconv2_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block8_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block8_sepconv3_act (Activatio  (None, 14, 14, 728)  0          ['block8_sepconv2_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block8_sepconv3 (SeparableConv  (None, 14, 14, 728)  536536     ['block8_sepconv3_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block8_sepconv3_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block8_sepconv3[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " add_30 (Add)                   (None, 14, 14, 728)  0           ['block8_sepconv3_bn[0][0]',     \n",
      "                                                                  'add_29[0][0]']                 \n",
      "                                                                                                  \n",
      " block9_sepconv1_act (Activatio  (None, 14, 14, 728)  0          ['add_30[0][0]']                 \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block9_sepconv1 (SeparableConv  (None, 14, 14, 728)  536536     ['block9_sepconv1_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block9_sepconv1_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block9_sepconv1[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block9_sepconv2_act (Activatio  (None, 14, 14, 728)  0          ['block9_sepconv1_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block9_sepconv2 (SeparableConv  (None, 14, 14, 728)  536536     ['block9_sepconv2_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block9_sepconv2_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block9_sepconv2[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " block9_sepconv3_act (Activatio  (None, 14, 14, 728)  0          ['block9_sepconv2_bn[0][0]']     \n",
      " n)                                                                                               \n",
      "                                                                                                  \n",
      " block9_sepconv3 (SeparableConv  (None, 14, 14, 728)  536536     ['block9_sepconv3_act[0][0]']    \n",
      " 2D)                                                                                              \n",
      "                                                                                                  \n",
      " block9_sepconv3_bn (BatchNorma  (None, 14, 14, 728)  2912       ['block9_sepconv3[0][0]']        \n",
      " lization)                                                                                        \n",
      "                                                                                                  \n",
      " add_31 (Add)                   (None, 14, 14, 728)  0           ['block9_sepconv3_bn[0][0]',     \n",
      "                                                                  'add_30[0][0]']                 \n",
      "                                                                                                  \n",
      " block10_sepconv1_act (Activati  (None, 14, 14, 728)  0          ['add_31[0][0]']                 \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block10_sepconv1 (SeparableCon  (None, 14, 14, 728)  536536     ['block10_sepconv1_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block10_sepconv1_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block10_sepconv1[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block10_sepconv2_act (Activati  (None, 14, 14, 728)  0          ['block10_sepconv1_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block10_sepconv2 (SeparableCon  (None, 14, 14, 728)  536536     ['block10_sepconv2_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block10_sepconv2_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block10_sepconv2[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block10_sepconv3_act (Activati  (None, 14, 14, 728)  0          ['block10_sepconv2_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block10_sepconv3 (SeparableCon  (None, 14, 14, 728)  536536     ['block10_sepconv3_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block10_sepconv3_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block10_sepconv3[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " add_32 (Add)                   (None, 14, 14, 728)  0           ['block10_sepconv3_bn[0][0]',    \n",
      "                                                                  'add_31[0][0]']                 \n",
      "                                                                                                  \n",
      " block11_sepconv1_act (Activati  (None, 14, 14, 728)  0          ['add_32[0][0]']                 \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block11_sepconv1 (SeparableCon  (None, 14, 14, 728)  536536     ['block11_sepconv1_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block11_sepconv1_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block11_sepconv1[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block11_sepconv2_act (Activati  (None, 14, 14, 728)  0          ['block11_sepconv1_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block11_sepconv2 (SeparableCon  (None, 14, 14, 728)  536536     ['block11_sepconv2_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block11_sepconv2_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block11_sepconv2[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block11_sepconv3_act (Activati  (None, 14, 14, 728)  0          ['block11_sepconv2_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block11_sepconv3 (SeparableCon  (None, 14, 14, 728)  536536     ['block11_sepconv3_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block11_sepconv3_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block11_sepconv3[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " add_33 (Add)                   (None, 14, 14, 728)  0           ['block11_sepconv3_bn[0][0]',    \n",
      "                                                                  'add_32[0][0]']                 \n",
      "                                                                                                  \n",
      " block12_sepconv1_act (Activati  (None, 14, 14, 728)  0          ['add_33[0][0]']                 \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block12_sepconv1 (SeparableCon  (None, 14, 14, 728)  536536     ['block12_sepconv1_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block12_sepconv1_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block12_sepconv1[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block12_sepconv2_act (Activati  (None, 14, 14, 728)  0          ['block12_sepconv1_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block12_sepconv2 (SeparableCon  (None, 14, 14, 728)  536536     ['block12_sepconv2_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block12_sepconv2_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block12_sepconv2[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block12_sepconv3_act (Activati  (None, 14, 14, 728)  0          ['block12_sepconv2_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block12_sepconv3 (SeparableCon  (None, 14, 14, 728)  536536     ['block12_sepconv3_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block12_sepconv3_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block12_sepconv3[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " add_34 (Add)                   (None, 14, 14, 728)  0           ['block12_sepconv3_bn[0][0]',    \n",
      "                                                                  'add_33[0][0]']                 \n",
      "                                                                                                  \n",
      " block13_sepconv1_act (Activati  (None, 14, 14, 728)  0          ['add_34[0][0]']                 \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block13_sepconv1 (SeparableCon  (None, 14, 14, 728)  536536     ['block13_sepconv1_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block13_sepconv1_bn (BatchNorm  (None, 14, 14, 728)  2912       ['block13_sepconv1[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block13_sepconv2_act (Activati  (None, 14, 14, 728)  0          ['block13_sepconv1_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block13_sepconv2 (SeparableCon  (None, 14, 14, 1024  752024     ['block13_sepconv2_act[0][0]']   \n",
      " v2D)                           )                                                                 \n",
      "                                                                                                  \n",
      " block13_sepconv2_bn (BatchNorm  (None, 14, 14, 1024  4096       ['block13_sepconv2[0][0]']       \n",
      " alization)                     )                                                                 \n",
      "                                                                                                  \n",
      " conv2d_11 (Conv2D)             (None, 7, 7, 1024)   745472      ['add_34[0][0]']                 \n",
      "                                                                                                  \n",
      " block13_pool (MaxPooling2D)    (None, 7, 7, 1024)   0           ['block13_sepconv2_bn[0][0]']    \n",
      "                                                                                                  \n",
      " batch_normalization_11 (BatchN  (None, 7, 7, 1024)  4096        ['conv2d_11[0][0]']              \n",
      " ormalization)                                                                                    \n",
      "                                                                                                  \n",
      " add_35 (Add)                   (None, 7, 7, 1024)   0           ['block13_pool[0][0]',           \n",
      "                                                                  'batch_normalization_11[0][0]'] \n",
      "                                                                                                  \n",
      " block14_sepconv1 (SeparableCon  (None, 7, 7, 1536)  1582080     ['add_35[0][0]']                 \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block14_sepconv1_bn (BatchNorm  (None, 7, 7, 1536)  6144        ['block14_sepconv1[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block14_sepconv1_act (Activati  (None, 7, 7, 1536)  0           ['block14_sepconv1_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " block14_sepconv2 (SeparableCon  (None, 7, 7, 2048)  3159552     ['block14_sepconv1_act[0][0]']   \n",
      " v2D)                                                                                             \n",
      "                                                                                                  \n",
      " block14_sepconv2_bn (BatchNorm  (None, 7, 7, 2048)  8192        ['block14_sepconv2[0][0]']       \n",
      " alization)                                                                                       \n",
      "                                                                                                  \n",
      " block14_sepconv2_act (Activati  (None, 7, 7, 2048)  0           ['block14_sepconv2_bn[0][0]']    \n",
      " on)                                                                                              \n",
      "                                                                                                  \n",
      " flatten_1 (Flatten)            (None, 100352)       0           ['block14_sepconv2_act[0][0]']   \n",
      "                                                                                                  \n",
      " dense_1 (Dense)                (None, 5)            501765      ['flatten_1[0][0]']              \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 21,363,245\n",
      "Trainable params: 501,765\n",
      "Non-trainable params: 20,861,480\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "55848fa0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 1s - loss: 1.9193 - accuracy: 0.3600 - 1s/epoch - 274ms/step\n",
      "Restored model, accuracy: 36.00%\n"
     ]
    }
   ],
   "source": [
    "# ResNet50 model\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "333ce1db",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 2s - loss: 1.9222 - accuracy: 0.3800 - 2s/epoch - 534ms/step\n",
      "Restored model, accuracy: 38.00%\n"
     ]
    }
   ],
   "source": [
    "# ResNet50 model while checkpointing\n",
    "model=load_model('best_resnet_model.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5c2ca704",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWoAAAD8CAYAAABekO4JAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAA6NElEQVR4nO3dd3hU1dbA4d9OIxUSIHQCAektQCiKUkSl2VCugmBX7O3a9btXvfau2FGxi0qzIBaQEkBUEnoJPZAQSCW9zsz+/tgTasqkDDPJrPd58iSZOXNmnSnr7LOr0lojhBDCfXm5OgAhhBCVk0QthBBuThK1EEK4OUnUQgjh5iRRCyGEm5NELYQQbs6hRK2UukcptUUptVUpda+TYxJCCHGcKhO1Uqo3cDMwGOgHXKiU6uLswIQQQhiOlKh7AH9prQu01hZgBTDRuWEJIYQo4+PANluAZ5VSzYBCYDwQW9kDmjdvrjt27Fj76IQQwkPExcWla63Dy7uvykSttd6ulHoRWAzkARsBy8nbKaWmA9MBIiIiiI2tNJcLIYQ4jlJqf0X3OdSYqLX+WGs9QGs9HMgEdpWzzUytdbTWOjo8vNyTghBCiBpwpOoDpVQLrXWqUioCuAw407lhCSGEKONQogbm2euoS4E7tNZHnBiTEEKI4ziUqLXW5zg7ECFE/VBaWkpSUhJFRUWuDqVe8vf3p127dvj6+jr8GEdL1EIIAUBSUhIhISF07NgRpZSrw6lXtNZkZGSQlJREZGSkw4+TIeRCiGopKiqiWbNmkqRrQClFs2bNqn01IolaCFFtkqRrriavnVsl6hl/7GLFzjRXhyGEEG7FrRL1Byv2sGKHJGohRPmysrJ49913a/TY8ePHk5WV5fD2Tz75JK+88kqNnquuuVWiXuV1M2cfqNmbIIRo+CpL1FartdLHLlq0iNDQUCdE5Xxulai90HiV5Lk6DCGEm3rkkUfYs2cPUVFRPPjggyxfvpxRo0Zx1VVX0adPHwAuvfRSBg4cSK9evZg5c+bRx3bs2JH09HQSEhLo0aMHN998M7169eKCCy6gsLCw0ufdsGEDQ4cOpW/fvkycOJEjR8xQkhkzZtCzZ0/69u3L5MmTAVixYgVRUVFERUXRv39/cnNza33cbtU9r9jLH29LvqvDEEI46KmftrItOadO99mzTWOeuKhXufe98MILbNmyhQ0bNgCwfPly/vnnH7Zs2XK0u9usWbNo2rQphYWFDBo0iMsvv5xmzZqdsJ9du3Yxe/ZsPvzwQ6644grmzZvHtGnTKozpmmuu4a233mLEiBH897//5amnnuKNN97ghRdeYN++fTRq1Ohotcorr7zCO++8w7Bhw8jLy8Pf37/Wr4lblahLvALwsRS4OgwhRD0yePDgE/okz5gxg379+jF06FASExPZteuUqYmIjIwkKioKgIEDB5KQkFDh/rOzs8nKymLEiBEAXHvttcTExADQt29fpk6dypdffomPjyn3Dhs2jH//+9/MmDGDrKyso7fXhluVqEu8A/GxSqIWor6oqOR7OgUFBR39e/ny5SxZsoQ1a9YQGBjIyJEjy+2z3KhRo6N/e3t7V1n1UZGff/6ZmJgYfvzxR55++mm2bt3KI488woQJE1i0aBFDhw5lyZIldO/evUb7L+NWJWqLdyB+tpq9YEKIhi8kJKTSOt/s7GzCwsIIDAwkPj6ev/76q9bP2aRJE8LCwli5ciUAX3zxBSNGjMBms5GYmMioUaN46aWXyMrKIi8vjz179tCnTx8efvhhoqOjiY+Pr3UMblWitvoE0siW6eowhBBuqlmzZgwbNozevXszbtw4JkyYcML9Y8eO5f3336dv375069aNoUOH1snzfvbZZ9x6660UFBTQqVMnPvnkE6xWK9OmTSM7OxutNffddx+hoaH85z//YdmyZXh7e9OzZ0/GjRtX6+dXWus6OIwTRUdH65osHLD1rSsISt9Ix6d21HlMQoi6sX37dnr06OHqMOq18l5DpVSc1jq6vO3dquoDvyACKaLYUnl/SCGE8CRulai1r0nUeUWnrPQlhBAey60StWoURCDF5BeVujoUIYRwGw4laqXUfUqprUqpLUqp2Uqp2vfgLoe3fwheSpOfX/uRPEII0VBUmaiVUm2Bu4ForXVvwBuY7IxgfPyDASjKr9uRTkIIUZ85WvXhAwQopXyAQCDZGcH4BIQAUCyJWgghjqoyUWutDwKvAAeAQ0C21vp3ZwTjZ0/URQWSqIUQdSM4OLhat7sjR6o+woBLgEigDRCklDpl9hKl1HSlVKxSKjYtrWZzSjcKbAxAaaHUUQshRBlHqj7OA/ZprdO01qXAfOCskzfSWs/UWkdrraPDw8NrFIx/sEnU1iJJ1EKIUz388MMnzEf95JNP8uqrr5KXl8fo0aMZMGAAffr04YcffnB4n1prHnzwQXr37k2fPn349ttvATh06BDDhw8nKiqK3r17s3LlSqxWK9ddd93RbV9//fU6P8byODKE/AAwVCkVCBQCo4HqDzt0gL+9RG0plDmphagXfnkEDm+u23226gPjXij3rsmTJ3Pvvfdy++23A/Ddd9/x66+/4u/vz4IFC2jcuDHp6ekMHTqUiy++2KH1CefPn8+GDRvYuHEj6enpDBo0iOHDh/P1118zZswYHn/8caxWKwUFBWzYsIGDBw+yZcsWgGqtGFMbVSZqrfXfSqm5wDrAAqwHZlb+qJrxtvf6sBVLohZCnKp///6kpqaSnJxMWloaYWFhREREUFpaymOPPUZMTAxeXl4cPHiQlJQUWrVqVeU+V61axZQpU/D29qZly5aMGDGCtWvXMmjQIG644QZKS0u59NJLiYqKolOnTuzdu5e77rqLCRMmcMEFF5yGo3ZwUiat9RPAE06OBfxMotaSqIWoHyoo+TrTpEmTmDt3LocPHz66qspXX31FWloacXFx+Pr60rFjx3KnNy1PRfMdDR8+nJiYGH7++WeuvvpqHnzwQa655ho2btzIb7/9xjvvvMN3333HrFmz6uzYKuJWIxPxs88rWyqrvAghyjd58mS++eYb5s6dy6RJkwAzvWmLFi3w9fVl2bJl7N+/3+H9DR8+nG+//Rar1UpaWhoxMTEMHjyY/fv306JFC26++WZuvPFG1q1bR3p6Ojabjcsvv5ynn36adevWOeswT+BW05zi448VL7xKJFELIcrXq1cvcnNzadu2La1btwZg6tSpXHTRRURHRxMVFVWtifonTpzImjVr6NevH0opXnrpJVq1asVnn33Gyy+/jK+vL8HBwXz++eccPHiQ66+/HpvNBsDzzz/vlGM8mVtNcwpQ8FRrlgWcz4SHPq/jqIQQdUGmOa29+j3NKVDsFYC3RVZ5EUKIMm6XqEu9A/GVdROFEOIoN0zUAfhJohbCrTmjytRT1OS1c7tEbfUJwk9L1YcQ7srf35+MjAxJ1jWgtSYjIwN//+rNFO1evT4Am08gATqLEosNPx+3O48I4fHatWtHUlISNZ3Tx9P5+/vTrl27aj3G/RK1XxBBFJFfbMHPx8/V4QghTuLr60tkZKSrw/Ao7ldk9QsiUBWTVyzrJgohBLhholZ+wQRRJIlaCCHs3C5Re/sHE0iRLHArhBB2bpmofZSN/ELpoieEEOCGidrH374cl6ybKIQQgBsmar9Ak6hLZN1EIYQA3DFRB5hVXkoKZDkuIYQAN0zU/kGmRC3LcQkhhOHIKuTdlFIbjvvJUUrd66yAvO111JYiqfoQQghwbM3EHUAUgFLKGzgILHBaRPZVXmzFsniAEEJA9as+RgN7tNaOr3NTXbJuohBCnKC6iXoyMLu8O5RS05VSsUqp2FpN1lK2bqIsxyWEEEA1ErVSyg+4GJhT3v1a65la62itdXR4eHjNI7InaiUL3AohBFC9EvU4YJ3WOsVZwQDgGwiAlyRqIYQAqpeop1BBtUed8vKmWPnjbZEh5EIIAQ4maqVUIHA+MN+54Ril3gGybqIQQtg5tHCA1roAaObkWI4q9Q7Et0SW4xJCCHDDkYkAFu9A/HUhFqvN1aEIIYTLuWWitvkGmjmpi62uDkUIIVzOTRN1EEGqmNxiWTxACCHcMlHjFyQlaiGEsHPLRK38gglSReRJiVoIIdwzUXv5mxJ1npSohRDCPRO1t38IQRSTLyuRCyGEeyZqH/9gGqlS8gukL7UQQrhloi5bjqtYluMSQgj3TNS+ZQvcFkqiFkIIt0zUPvbluEolUQshhHsm6rI5qa1FssqLEEK4eaKWErUQQrh1otayHJcQQrhropYFboUQooybJmpZN1EIIco4usJLqFJqrlIqXim1XSl1plOjsidqr1JZ5UUIIRxa4QV4E/hVaz3Jvhp5oBNjAl+TqGXdRCGEcCBRK6UaA8OB6wC01iVAiXOj8sOifPGRRC2EEA5VfXQC0oBPlFLrlVIfKaWCnBwXpd6B+NkKsNq0s59KCCHcmiOJ2gcYALynte4P5AOPnLyRUmq6UipWKRWblpZW68AsPgEEqWLyZAY9IYSHcyRRJwFJWuu/7f/PxSTuE2itZ2qto7XW0eHh4bUOzOZTtsqLJGohhGerMlFrrQ8DiUqpbvabRgPbnBoV9nUTKZIStRDC4zna6+Mu4Ct7j4+9wPXOC8nON4hAlSGJWgjh8RxK1FrrDUC0c0M5SaMggkgmQxK1EMLDuefIRMCrUbBZN7FIErUQwrO5baL29i9biVwStRDCs7lxog6xr0QuiVoI4dkcbUw87XwDgvFRxeQXOXcQpBBCuDu3LVGXLcdVXCgz6AkhPJvbJuqyGfRKC3NcHIgQQriWGydqs3iArJsohPB0bpyoTYnaUiiJWgjh2dw+UctyXEIIT+fGidpUfVAqiVoI4dncOFHbp7yWlciFEB7O7RO1LHArhPB0bpyoTdWHjyxwK4TwcG6cqE2J2sdaiE2W4xJCeDD3TdQ+/tjwIlAVkV8i830IITyX+yZqpbD4BMoqL0IIj+fQpExKqQQgF7ACFq31aVlEwOoTKOsmCiE8XnVmzxultU53WiTlsPkG2eektp7OpxVCCLfivlUfYNZNpFhWeRFCeDRHE7UGfldKxSmlpjszoBM0CpJVXoQQHs/Rqo9hWutkpVQLYLFSKl5rHXP8BvYEPh0gIiKiToJTfsEEkUKSJGohhAdzqESttU62/04FFgCDy9lmptY6WmsdHR4eXifBefsHE0ixNCYKITxalYlaKRWklAop+xu4ANji7MBAFrgVQghwrOqjJbBAKVW2/dda61+dGpWdt3+I9KMWQni8KhO11nov0O80xHIK5RdsRiYWlbri6YUQwi24d/c8vyC80RQXysRMQgjP5eaJ2sygV1qU6+JAhBDCddw8UZsZ9GyyHJcQwoPVj0QtK5ELITyYmydqU/WhSyRRCyE8l5snavtyXLJuohDCg9WPRC3LcQkhPFi9SNTelnxZjksI4bHcPFGbOupAiigolTmphRCeyc0TtSlRyyovQghP5t6J2jcQgCBVLPN9CCE8lnsnai8vLPZ1E2WVFyGEp3LvRA3Y7CuRS9WHEMJTuX2i1n5BBKoiciVRCyE8lNsnavyCCJJVXoQQHsztE7WXX7D0+hBCeDT3T9T25bik6kMI4akcTtRKKW+l1Hql1EJnBnQyr0ZSohZCeLbqlKjvAbY7K5CKKL9ggr2KpXueEMJjOZSolVLtgAnAR84Npxx+QfYFbmUIuRDCMzlaon4DeAiwVbSBUmq6UipWKRWblpZWF7EZfkEESNWHEMKDVZmolVIXAqla67jKttNaz9RaR2uto8PDw+ssQPyC8cNCQZFMdSqE8EyOlKiHARcrpRKAb4BzlVJfOjWq49knZkpJyzxtTymEEO6kykSttX5Ua91Oa90RmAws1VpPc3pkZeyJOicnm9ScotP2tEII4S7cvh/10alOVRHrE7NcG4sQQrhAtRK11nq51vpCZwVTLvviAU28i1l/IOu0PrUQQriDelOi7tnUiw2JR1wcjBBCnH71JlH3aO7N5qRsrLJ2ohDCw9SDRG2qPrqGKfJLrOxKzT12X/pu2DLPRYEJIcTpUQ8StSlRRzZWAMfqqfMz4ItLYd5NUCq9QYQQDVe9SdTNfEsIDfRlw4EssFlh/k2QnQjaBpl7XRujEEI4Ub1J1Kq0gKj2oWxIzILlz8OepTDoJrNN+g7XxSeEEE7m/ona2xe8G0FJHlHtQ2mfvhxiXob+V8P5T5tt0ne5NEQhhHAm90/UYErVJfmcGZrNaz7vkde0N4x/BfwCoUkEpO90dYRCCOE09SRRB0N+GtF/340VL+Z3eR58/c19zbtAmlR9CCEarnqSqINg2w94p23n+YD7WZkWeOy+8G6QsRtsFc7AKoQQ9Vr9SdQAox7HEnkuGxKz0No+8KV5FygtgJyDrotPCCGcqH4k6shzoO9kOOd+oiJCScstJjnb3ne6eVfzW+qphRANlI+rA3DIeU8e/TOqfSgA6w8coW1owImJ+ozRpz82IYRwsvpRoj5O91aN8fPxMgNfAILCwT9UStRCiAar3iVqPx8v+rRtYga+AChlStXSl1oI0UDVu0QNpvpj88FsSq32nh7Nu0oXPSFEg+XI4rb+Sql/lFIblVJblVJPnY7AKhPVPpRii434Q/aZ9MK7Qn4qFMp81UKIhseREnUxcK7Wuh8QBYxVSg11alRVKGtQPLqQwNEGxd2uCUgIIZzIkcVttdY6z/6vr/3HpbP3twsLoHmw37E1FKWLnhCiAXOojlop5a2U2gCkAou11n+Xs810pVSsUio2LS2tjsM85bmIah92rEExtAN4+8ksekKIBsmhRK21tmqto4B2wGClVO9ytpmptY7WWkeHh4fXcZin6h8Ryt60fLILSsHbB5p2lp4fQogGqbqrkGcBy4GxzgimOo7WUydlmRuad5GqDyFEg+RIr49wpVSo/e8A4Dwg3slxValvuyYoxbGBL827QuY+sJS4NC4hhKhrjpSoWwPLlFKbgLWYOuqFzg2raiH+vnRpEXys50d4N9BWWZZLCNHgVDnXh9Z6E9D/NMRSbVHtQ/ltawqFJVYCmncxN6bvhBbdXRuYEELUoXo5MrHMv6Lbk11YynvLd0Oz4xK1EEI0IPU6UQ/q2JSL+7Xh/Zi9JOZ7QeN2kqiFEA1OvU7UAI+N74GPl+KZn7dJzw9x+mkN67+E2FmwdwVkJ8lqQw2B1vD3B5Cxx9WRAPVlPupKtGrizx2jzuDl33ZwsF972iatNS+yUq4OTXiC/avhhztOvM3HH8IioVlniL4ezjjPNbGJmkvZAr88BL0mwr8+dXU09T9RA9x0TiTfxSYyJyGQe0vyIPcQNG7j6rCEJ4h5GYJawA2/mtJ05h5TCsvcZ5J4dpIk6vpo07fmd/wiM9lbQJhLw6n3VR8AjXy8+e+FPfk7t5m5QaY8FadD4lrYuxzOusuUnjuNgOgbYMyzMOVrGHIbHNoIBZmujlRUh80Km+eZ0c7WYti6wNURNYxEDXBu9xa0iOwDQO7B7S6ORniEmJdNSSv6hvLv7zQS0JCwsnbPozXs/9MkEOF8+1dDbjKMegzCe8CG2a6OqOEkaqUU91xyNjk6kE0b/nF1OOJ4hzdDYZaro6hbhzbCrt9g6B3QKLj8bdoOAL9gU+qujZ2/wSfjYPOc2u3HE9msZg6grQtg0xxz0qvKpm/N+9ZtPERNgaR/XD6FcoOooy7TqUUIB4MjIW0nGxKzjs4HIlzEUgJ/PAVr3oZ2g+H6X8wEWg1BzMvQqAkMmV7xNt6+0PFs0xukNv56x/yOXwj9JtduXw1dTjJs/8k0Bh7eAqnbwVJ47P7gcPuVTgVKi2Dbj9DjIvALhL5XwpInYeNsGP0fZ0dfoQZToi7TolMfungf4okft2KzuXTabM+WuQ8+GWuSdOdzTalk5SuujqpupG43yWDIdPBvUvm2kSNMA2NWYs2eK2Ur7Isxz7N7qUkkonyZ++DDc01vje0LzZVO9PVwybtw81IIaQPLnq+8VL3rNyjOgT7/Mv+HtDKf343fuLTbZYNL1L4tutGCTPYkJrNw8yFXh9PwpGyDuTeafsO5KeVvs2U+fDDcXC5e8TlcvcCUTFa8BIkNoFpq5avgGwRDb69627LS274alqr/eg98A2H8K1CaDwmrHHtc4REozq3Zc9ZH2Unw+cVgKYLpy+GhvXDtTzD2eeg/FdoOhHP+DYl/wd5lFe9n03cQ3NKcYMv0mwI5SZAQ4/TDqEiDS9SEdwPg3ObZvPb7jmML4NZEUizMuwl+fbT2cRVkwryb63eiKsqGb64y9X0L74NXu8HHF8DqGWYyrJIC+OkemHu9eR9uXQk9LzGPHf8yNGlrXs+iHNceR21k7IEt82DQjRDYtOrtW/Qw3fdqUk+dn24SR78p0ONic3LYsajqx2kNn18C7wyF3MPVf976Ji/VHG9hlikUtOlf/jiKAdeY0csVlaoLj8Cu36H35SdW0XWfYKq5XNio2PAStX1Zrlt6WkjIKGBeXNKx+4pyTGko7jNT2ivvzbJaYOv38NH58NFo8/df78KOX2seU2EWfDERNn9nBkdYS2u+L1fR2sSedQCuXwS3rTGt4qWFsPg/MKM/vNod4j6FYfea+uiwDsce798ELvsQshPhl4frNraiHHOC2FfL3hWOWPmaWU3ozDsd214p021v7wrHGrKOF/uJ6R425Fbw9YfOo2Dnr1XvJ3m9aezMSYLZk6Ekv3rPW58UZJoknZMMU+eYJF0Rn0Yw/H5TDbfnj1Pv3/YDWEuOVXuU8Q2AXpfC9h9ddpXS8BJ1WEfw8qGnXwpR7UOZ8ccuikqtpnT8/tnwx//gp7vh7YHwSlf49mr46304GAd/vm0SzpxrzarmY1+EB3ZC827wy4OmxFhdxbnw1b9MXePQ280Q99hPan+c1akvs9nMiSY/vebP99e7pl72/KcgYii07AkjHjKl5ns2wZjn4YxzYdo8s42376n7iBgKwx+EjV+bUmldsNlg/nRzgvhiomnZd5Yj+2HTNzDwOghp6fjjIkeYz1NqNbqNWkpg7YdmsEy4fU3QbuMh5yAc3lT5Yzd8ZUZHTpxpEvb86Q2za19RtnnPM/bAlNnm81WVqGnQJAKWPXfqCW/THGh2RvnJPuoqKC0wDY0u0PAStbcvNO2MSt/JQ2O6cSi7gC3fPgmzxpg35obf4Y61cNGbppEgeQP8+rBphPj9cQhtD1d+BXetg6G3msvbCa+akuSq16oXS0kBfD3ZnAQmzYIxz0HkcFj+vLnMqgmtTSv0CxGw/MWqTx6p8aZr1+wrYcEtNXvOA3/D4v9C9wvLL0mGdYAzbzdDbasahTf8IWg3yFSd1LSB7XjLnoGdv8Do/0L7ITD/Jlj1RvVLr45Y/Sag4Ky7q/e4snrq6lR/bF0AeSkw9LZjt3UdY55/xy8VP660yHTj634h9LvSnEDjF5r3ryEpyYevrjC9O678ovKeHMfz8TOl6oNxsGvxsduzEmH/KtOWUl61Sfsh0LST6f3hAg0vUcPRyZnOamnhpyavEL17BpauE0zpL2KIKaEMvA4u+wDu2wz3bYVJn8D0FeayvseF4OV9bH+R55g3cPWbjq/LaCmGb6eazvMTP4CeF5sPwAXPmiQdU4MeEFqb+vJVr0NoBCx/Dt6OLr9/aGkRLH3WXEWk74Cel8LuJSd+OB2Rn27qnJu0g0veqf0cKt4+cNlMU8JbcGvtSnqb55qqrIHXwdn/hqvnm7kZljxhWv7L23dpIaz9GN4aCG8PMldajkiKhfVfmIapJm2rF2doe/Mld7RBUWv4+z1zJdd59LHbg5qbhFFZPfWOn01Js/9U8//QW2HwdNP7Zu3H1YvbXVmKTVtJ0j9w+Uf2E1g1RE099v0p+95smWt+95lU/mOUMm0FCSvNldVp5shSXO2VUsuUUtuVUluVUvecjsBqpXlXczn03ln0tMTzcOnNvB/+fxAQWv72TdpB78ugTVTF+zz/afAJgEUPVF1as5bCnOtgz1K45G3oe1ydV+u+0H9a9Wfmstng5/vNF3jo7XDbalMPHNTclCI/Pv9Y0tm3Et4fBjEvmeO6M9bUDzftDL895ngduc0K8282yfqKzyt+/aqraSfTuLh/lamGSlhtvnzVkbwefrgTIs6CcS+bL5JPI7h8lin1/zMTvrvGJGYwx7D8BXi9F/z8b2gUYk5ms8aYEnhFVUmWEvjjafP6BrWAcx6o2TF3Gml6bDjy2if+bY5v6K2nnhi7jTXVGdkHy3/s+q+gSfsTey2MeR66jIFFD8KuJac+RmtzxXhwnXOuRMpkJZrP/aY5psth2k5zUqnOc9ps8P1t5urk4rfNibm6vH1NFVzyejOYCExM7QaZz2ZFyvqwl80Dcho5MvrAAtyvtV6nlAoB4pRSi7XW25wcW8216GGW5Qppjdf1s8hYlMuilfu4+sxImgSWU3fqiJCWpsP7ogfMZWnvy8rfzlJs6gR3LDJdqvpPO3Wbc/9j9rH4vzD5q6qf22Y1CW39l6ah7rwnzRe4w1lw83JzOfbHU6bxs90gSFpr6uqvXmCqd8qMedY0Lq392CSBqsS8Yk42F74BrftVvX119JtiesDEfWqOy8cf2g+GjueYQSJtB5rEW57cFPhmqjlJXfG5uZwt4+VljrNxW3NS+vxSaNnL1NtaiqDrWFN10eEsKMqCH+82JfB9K8yVT3CLY/s6vNmU+lO2mLrNsc9V3W+6Ip1Gmi6NB9eZq7rK/PUu+IdC33IGt3Qbb6q+dv5qep4cLzvJvF/DHzzxitDbByZ9DLPGmQLElNlQkmcS1cF15neBvf2i5yWmWrCuJyFK2Wrqk/PK6dLpE2A+r+NeNA2vFdHaVE9umWe+A2VXDTXRb4q5Glv+nCmopW41J/zKhEaYz+fG2eY1Po0zdCpdzTOoUuoH4G2tdYXX0NHR0To21sFLSmewlpqGr27jwdef7YdyGD9jJbeN6MxDY2uxTJfNCh+OMonizrXg3/jE+3cvMT0aMnabKo6zKukZEPMKLH3a9PWMHF7JsVjgh9vNWXzEwzDy0fI/IMV5pkpk4zemBD/8ITOy6nhamy9L8nq4e33l3ct2LYGvJpkqn4nvO+9DWXgE9q8xl5QJK81oMrTpitZ5lHkPu44xSRnMifDTC03yvOE3c4VSka0LYP4tZn99rzSTJ9m7bx6lNcR9YqqUGjU21WEdh8Pq100bQGBTuGiGKcnWRkEmvNTJvH8jK+n1knUA3uwHw+4xyehkWsNbA8zV0bS5J94X8zIsfQbu3gBNI099bPZBczLPtY8vUF4Q3t00nrXpb96LFS9CSGu4/OOqTyiOSoqFLy83/cEnfwV+QabbYF7Ksd87fzMDg87/n7kiKu/ztnqG6WE05FYY+0LtP5PrvzLfrdb9zOfugZ3HPmcV2fC1KdHf8JtjjZfVoJSK01pHl3un1trhH6AjcABoXNl2AwcO1O7m7tnrdPf/+0Wn5BSecp/NZtMJ6Xm6qNRS9Y4SY7V+oonWvzx67LbMBK1nX6X1E421frO/1rsWV72fkgKtX+ul9XvDtLZW8LyWEq2/u9bsd8VLVe/TEYe3av1kqNY/P1jxNrsWa/10C63fPUvr4ry6eV5H5WdovX2h1j/dq/Ur3c2xP9FE64/O13rla1rPvdHctmW+Y/vL2KN1bkrV2x3eqvXbg81zvdnfPMd315l46sr7w7WeNa7ybX57XOsnw7TOSqp4m18e1fp/zbUuyj12m82m9ZtRWs8aX/n+M/Zq/c+HWif8eeLjyySu1fr1PiaG5S9V/Nl01J5lWj/TWus3+pnvSUWKcrT+9mrzus+5/tTP3cZv7e/JtVpbrbWLqYyl1LxmTzTW+stJjj2mKNcczzfTzGteh4BYXVHureiOUzaEYCAOuKyC+6cDsUBsREREnR5AXdiXlqc7PfqzfuKHLVprrbPyS/TCjcn6oTkb9dDnlugODy/UY15foTPziqve2Y/3mA9yYqzWy180Se2ZVlrHvKp1aZHjQW2aYz4kcZ+feHtmgtYrXtb6rWhz/6o3Hd+nI366z8SfGn/qffGLTBJ4b5jWeel1+7zVZbNpfXC91kuf0/r9c+xJu7HWfzzjnOcrztf6x7u1fqWb1pvn1v3+f/+P1k81q/jkV5Sr9XPtzQmiMntjzOuw9YdjtyWsNret/6r2cRZmaT3nBrO/TyZonX2wZvvZvtB8lt4ZqnXOoaq3t9nMyfjJUK3fOdOcZLXWetcSrZ9qamKpzvfLERtmm+PcNMfxxyx/yTxmyVN1Gkplidqhqg+llC+wEPhNa11lHzWXV31U4NH5m5gbl0Sftk3YkJiFTUNjfx/O7tKcHq0a89ay3XRpEczXNw2tvC67INP0tijIBLRp0LjgGVPXVR1am0aqrANw4++mR8bmOaYxCaD9UDOfRO/La3zM5crPMP3F2w8y/Z7LbPsB5t4Arfqa2x0ZeXc6ZSeZvsidR5u66Ppmz1JT9TR1HnQ5qRujzQYL74F1n8ONS8x7UxFrKbx8hhkxd+m75rbv74Bt35vLd7+g2seqtbnMX/SAaT/oeTG0GWCqSFr0KL+f/PE2fgPf3262nzqnep+l3X/AvBtB22DEI7DsWVOHff2imrcRVERrOPCXqcZwtCpFa1h4r2lfGfuiY+09Dqis6qPKRK2UUsBnQKbW+l5HntBdE/Wh7EImvvMnLZv4M6JrOCO6Nqdfu1B8vM2XftmOVG75PI4erUP44qYhNPav5MO4/SczD8OIhxzvw1mexLXw8XFf2vAepo6596QTR/bVtTXvmMa2qXOhy/mm1XvBLdAu2nyx6voLIUyf9xc7mO5yY549drvNCj/eZRo8z7nf9Amvyrybzei6B3aZni2vdDUN3Je8Xbcxp++C3/8PDqwxPTTAJO5WfUwS9g81o/nKfizFZpDXtu9N28vk2RVPA1uZIwnwzTRI2WwGqNz4OzRuXYcHVks2q+lVFL/Q1OdX1K2vGmqbqM8GVgKbgbI+TI9prSvszOmuidoRS7alcNtXcfRu24TPbxhMSGXJuq6sngEFGWboastep6c12VIC751pGpTOvAN+utf0tpjyTc2+WMIxn15oepvcap9cyWoxjVObvzMNjSMeduz93zLPXP3c8Jvp5vnD7U5p4DpKazOfS/L6Y71FDm00E0V5+4F3I1PK9rH/7nA2XPi6GfpeUyUFpptlz4sr7zbnKqVFppE08W+46ls4Y3TVj6lErRJ1TdTnRA3w65bD3PH1Ovq3D+WzGwYT1KiBzKF8sh2/mhGLAJ1GweSvT+0pIupWWc+MB/eYq5b5N5veKaP/a0rTjirKNr1IzrzDXJXlp5r+8qdzUeey3OHJC0kXZcMnE8xJ7NqfoN3AGu+qskRdDyv6nG9s71bMmNyf9YlZ3PDpWgpKLK4OyTm6jjF9dXtfbkrSkqSdr9Mo83v3H6ZP89YFpn2jOkkaTJLveLaZ0e3An2YuitOdMJXy7CQN5n2YNtd06/tqkuMjl6tJEnUFJvRtzWtX9GNtQiZTP/qb3al5rg6pxootVr755wBJR06aF0Qp02940qzaXaIKx7WOMv21f7rb1G+Oe8n0766JruNMSVp5mQEcwjVCWpnBZV7e8MVlZkxDHZNEXYlLotry9lUD2JuWz/g3V/LGkp0UW+rXLGT/7Mtk/JsreWT+Zh5fsMXV4ZzAYrVh9bRVeLx9TCObpcjU4Q6p4URZcGwQTufR0LhN3cQnaqZZZ9Mwf86/ndLGI3XUDkjLLebphdv4cWMyncODeP6yvgyOdLOuayfJLizlhV/imf3PAdqGBjCwQxg/bkzmpzvPpk871/To0FqTkFHAyl1pxOxMY82eDHq2acy308/Ey8uDLqGzD5quhnUx8u+v900VSKvetd+XcClpTKwjy3ek8n/fbyHpSCFTBkfwyLjuNAk4Db1CqkFrzaLNh3nyp61k5BVz49mR3Hd+V6w2zbAXlnJm52Z8cHX5o1SdZeWuNH7ZcpiYnWkkHTGTJLVvGkCXFiEsjU/ltSv6cdmAavZBF6KBqSxRN9DuDM4xslsLfr9vOK8v3snHq/axND6F96YNZEBE7SawycgrxtfHq/J+25XQWrM3PZ91+4/w8+ZDLN+RRu+2jfnkukH0bnus9HzdsEhm/LGLHYdz6dYqpFYxOyKroIQnftzKDxuSCW7kw5mdm3HL8E4M7xpOh2ZB2GyaS99dzYu/xjO2dysC/Wr+cdRak1tsISOvhIy8YtLzSsjML+FIQQnDu4S77CpCiLogJeoa2pyUzR1fr+NwdhHPTOzNFdHta7Sf2IRMbvh0LTYN04Z24KZzImkeXMGscXZFpVZiE46w7oD5WX8gi+xCM31mkwBf7hx1BtcP63h0IE+ZI/klnP3iUkb3aMmMKZUsWVQHlsan8Mi8zWTml3DXuV24bWRn/HxObRKJTchk0vtruHt0F/59ftcaPVdiZgFXfLCGQ9nlr9DdyMeLt68awPk9q7EqixCnmVR9OElWQQl3fr2eVbvTuX5YRx4f3+OU5FiZP7ancPtX62gbGkCPNo1ZtPkQjXy8mDwogltGdKJ1k4Cj2+YWlbJsRxq/bTnMsh2pFJSYRs0uLYIZEBHGgA6hDIgIo3N4cKX1vc//sp0PY/byx/0jiWxeB0ONT5JbVMrTC7fxXWwS3VqG8OoV/U4o1Zfnzq/XsXhbCksfGEnb0IBKtz2Z1aaZMvMvth3K4e7RZxAe0ohmQY1oGuRH8+BGKAXTv4hjc1IWL1zWlysG1eyEKoSzSaJ2IovVxnOL4pm1eh9ndW7GO1cNICzIr8rHzYtL4qF5m+jVxlRRNAtuxJ60PN5bvofv1x9EKbh8QDv6tQ9l8bYUVu1Kp8Rqo3mwH+f3bMUFPVsyoENYtevI03KLOfvFpVwS1YaXJtXtHNOrd6fz0NxNHMou5NYRnbnnvC408vGu8nFJRwoY/eoKxvRqVe2S/vsr9vDCL/G8+q9+XD6w/Hru/GILt34Zx8pd6Twyrju3DO+E8vT+vx5Ia+3W77sk6tNgTmwijy/YQssmjfjwmmi6t2pc4bYzY/bw3KJ4hp1hGvaCTxr5mHSkgA9W7OXb2ERKLDbahQUwplcrxvZuxYCIMLxr2UPiyR+38uVf+1n+4EjahdV+kEt+sYUXfonni7/206l5EK9e0Y/+1ay3f/X3Hby1dDfzbjuTgR0c61GzLTmHS95ZxXk9WvLu1AGVfglLLDbun7ORnzYmc9PZkTw2vodn9TTxcLNW7WNmzF4+v3EwXVs6v32mJiRRnybrDhzh1i/iyCkqZUhks6NVElHtQwnx90VrzQu/xPNBzN6jA2oqK3Gm5RaTmV9C15bBdVoSSM4qZMTLy5gyOIL/XVK7bl1r9mTw0LyNJB0p5IZhkTxwQTcC/KouRZ8sv9jCua8up1VjfxbcPqzKJFpUauWSt1eTWVDCb/cOp6kDVzE2m+apn7by2Zr9XNa/LS9O6otvNaqqRP0Utz+TKz74C6tN075pAN/fPoxmVbQDuYIk6tMoJaeIGX/sIjbhCDtTc9HaDADs2iKE0EBf/t6XydVDO/Dkxb1qXTKujUfmbWL++oOsemgULRpXf1RiQYmFF3+J57M1++nYLJCX/9WPQR1r17d8XlwS98/Z6FB3vecWbWdmzF4+uX4Qo7q1qHTb42mteXvpbl5dvJOBHcK4c9QZjOgaLqXrBiqroIQJM1bh5QXPTezDTZ/F0rddE768aYhD1XKnkyRqF8kpKmXDgSx774wsdh7OZeqQCO489wyX15Xtz8hn1CvLufHsSB6f0POE+yxWGxuTstifUUCIvy8h/j409velcYAPjQN82Z6cw4NzN5F4pIDrzurIQ2O616gUfTKbTTPx3dUczili2QMjK+yut2ZPBld99BdTh0TwzKV9avRcc+OSePm3eFJyiukcHsT1wyK5fEC7U44jOauQP7ansGR7KusPHKFLyxAGRzZlcMemNWojEKeP1ppbvohj2Y5U5t56Fv3ah/LjxmTunr2eSQPb8fKkvi7/Hh5PErUo133fbuDXLYdZ/ci5FJRYiNmZTszONFbvSSe3qPKJqCKaBvLypL4M6dSsTmOK25/J5e+t4ZYRnXjggm6nVE3kFJUy7o2V+Pl48fPdZ9eq73WJxcaizYf4eNU+Nh/MJjTQl6sGRzC8azh/7k5nyfZUth3KAaBjs0AGdWzK7rQ8NidlY7FplILurRozJLIpNwyLJKKZTGrlTj5dvY8nf9rG/03owU3nHJsm9fXFO3nzj108Mq47t47o7MIITySJWpRrV0ou578eQ2igL1kFph92myb+DO8azjldwunROoSCEis5haXkFJWSU2ghp6gUby/FlYPa1ypJVubu2ev5cWMyft5edG0VTM/Wjc1PmyZ8/fd+ftp0iLm3nlntBsuKaK2J3X+Ej1fu4/dth7Fp8FIwsEMYo3u05LweLekcHnS09FVQYmHDgSz+SchkbUImcfuPEOTnw8fXDSKqfWidxCRqZ8vBbC5790/O6dKcj66NPqHkrLXmrtnr+XnzId6fNpAxvVq5MNJjJFGLCr3wSzzxh3M4p4tZ8aZzeN02XNZEUamV37elsC05h63J2WxLziEjv+To/feM7sJ9NRwcU5UDGQVsO5TN4MhmDjVQAuxJy+PaWf+QkVfCu9MGVKvO3JP8vvUwcQeOcOeoM5y6IEduUSkXvbWKYouNRXefU2532aJSK1fO/ItdKbnMufVMerVx/cjV2q7wMgu4EEjVWjvURUAStahLWmvScovZmpxDWm4xEwe0dbveGqm5RVz/yVriD+fy/GV9Khypml1QynexiWQWlDB1SESddI90d6k5Rfz3h638uvUwAJ2aB/HutAGVdmGtKa01d3+zgZ83JfPN9DMrnTwtNaeIS95ZDcBtIzsT3aEp3VqF1LiRP25/JpuTsrluWGSNHl/bRD0cyAM+l0QtRMXyii3cZh9Yc//5XU9oNN6TlsenqxOYG5dEYakVby+FAi6OasNtIzrTpZK+vel55iTVq03jKqcXcCdaa76LTeTZn7dTZLFx73ld6Ns2lHu/3UBecSnPTexT55NxldVLP3BBV+48t0uV229NzubWL+NIzDSThYU08qF/hzAGdQgjumNT+keE4u9beUN5bEImbyzZxard6bRq7M/yB0dW+Zjy1LrqQynVEVgoiVqIypVYbDw8bxML1h9k6pAILujVik9W72P5jjT8fLy4NKoN1w+LJDTQlw9j9jH7nwMUlloZ06slt488g37tQ8kuLOWffZn8uSedP3dnsCMlFzDdPKM7hHFBz1ac37MlHas5BUCp1cbS+FQOZBRw1hnN6Nm6sdOquRLS83l0/mbW7M1gSGRTnr+sD53CzTzNqTlF3DV7PX/vy2TK4AieuKhnjRLb8bTWvLZ4J28t3c253Vvw4TXRDpeMtdYczCokNuEIaxMyT+haG+DrzbAzmjOqezijurWgzXFTHKxNyORNe4JuHuzHLcM7M3VoRI3bbiRRC3Eaaa156bcdvLd8DwDhIY24emgHrhoScUqJODO/hE9X7+PTPxPIKbLQsVkgBzILsGkzmdSgjk05s3Mz+rRtQtz+IyzelnK0J0rXlsGc37MlZ3VuTt92TSqs9z2QUcA3aw8wJy6JtNzio7e3bNyIkV1bMLJbOMO6NK/x7I0nH/snqxN48dd4/Ly9eGxCD66Mbn9KP3WL1cari3fy3vI99GrTmPemDqxxr5lii5WH5m7ihw3JXBndnmcm9q511Vh2QSmx+zNZsTONpfGpR6fn7d4qhBHdwtl6MOdogr51RGemDulQ6y6qpyVRK6WmA9MBIiIiBu7fv79m0QrRQPy4MRmrzcb4Pq2rHFyRW1TK138f4M89GfRrH8pZnZvRPyK03MclZhaweFsKi7el8E9CJlZ7V8EzwoOJah9Kv/ZmNOx+e4JeuSsdLwWjurVgyuAIerVtzKpd6SzfkUbMrjRyiyz4eCn62JO9n7fC19sLPx8vfL29CPD15oro9lVOFZtXbOHhuZv4efMhzuvRgmcn9qFlFYOplmxL4d/fbcCmYUyvVpzbvQXndHX8pHEkv4TpX8SyNuEID47pxu0jO9f5VYLWmt2peSzbkcrS+FRiE44QGuhbZwm6jJSohWigsgtL2ZiYxYbjfjKP6yHTNjSAKwe151/R7U6YjbFMqdXG+gNZLNuRStz+IxRbbJRabJRay340RwpKKCy1Mm1IBx64oBtNAk9NortTc7nlizj2pefz8NjuTK/GxFeJmQW8+vsOlu1II7uwFB8vxaCOTRndowWjuregU/OgcveVkJ7P9Z+u5WBWIa/8qx8X9zs9y5HlF1uOnsjqkiRqITyE1prEzEI2JGURGuDLsDOa13qqgpyiUl77fSefr0kgLNCPR8f34PIBbY8mz4Wbknlo7iYC/bx5a8oAzuxcs0FQFquN9YlZLI1PZen21KN18418vGgbFkC7sEDahQXQLiyAxv6+vPr7DgA+vCaa6FpOX+AOatvrYzYwEmgOpABPaK0/ruwxkqiFaHi2Jmfzn++3sO5AFoM6hvHERb2Yv+4gs1bvY2CHMN65agCtmtTdavZJRwpYuSudfen5JB0pIOlIIUlHCo9eMUQ2D+KT6wZVu1HVXcmAFyFEnbDZNHPiEnnhl3iO2EezXndWRx4b36POqwIqkl9sITmrkPZNA2vdW8SdyJqJQog64eWluHJQBGN6teK9FXvo1y6U8X1an9YYghr5VNrvvCGSRC2EqLbQQD8eHdfD1WF4DPcahyuEEOIUkqiFEMLNSaIWQgg3J4laCCHcnCRqIYRwc5KohRDCzUmiFkIINyeJWggh3JxThpArpdKAms5z2hxIr8Nw6gs5bs8ix+1ZHDnuDlrr8PLucEqirg2lVGxF490bMjluzyLH7Vlqe9xS9SGEEG5OErUQQrg5d0zUM10dgIvIcXsWOW7PUqvjdrs6aiGEECdyxxK1EEKI47hNolZKjVVK7VBK7VZKPeLqeJxJKTVLKZWqlNpy3G1NlVKLlVK77L/DXBljXVNKtVdKLVNKbVdKbVVK3WO/vaEft79S6h+l1Eb7cT9lv71BH3cZpZS3Umq9Umqh/X9POe4EpdRmpdQGpVSs/bYaH7tbJGqllDfwDjAO6AlMUUr1dG1UTvUpMPak2x4B/tBadwH+sP/fkFiA+7XWPYChwB3297ihH3cxcK7Wuh8QBYxVSg2l4R93mXuA7cf97ynHDTBKax11XLe8Gh+7WyRqYDCwW2u9V2tdAnwDXOLimJxGax0DZJ508yXAZ/a/PwMuPZ0xOZvW+pDWep3971zMl7ctDf+4tdY6z/6vr/1H08CPG0Ap1Q6YAHx03M0N/rgrUeNjd5dE3RZIPO7/JPttnqSl1voQmKQGtHBxPE6jlOoI9Af+xgOO2375vwFIBRZrrT3iuIE3gIcA23G3ecJxgzkZ/66UilNKTbffVuNjd5c1E1U5t0l3lAZIKRUMzAPu1VrnKFXeW9+waK2tQJRSKhRYoJTq7eKQnE4pdSGQqrWOU0qNdHE4rjBMa52slGoBLFZKxddmZ+5Sok4C2h/3fzsg2UWxuEqKUqo1gP13qovjqXNKKV9Mkv5Kaz3ffnODP+4yWussYDmmfaKhH/cw4GKlVAKmKvNcpdSXNPzjBkBrnWz/nQoswFTv1vjY3SVRrwW6KKUilVJ+wGTgRxfHdLr9CFxr//ta4AcXxlLnlCk6fwxs11q/dtxdDf24w+0laZRSAcB5QDwN/Li11o9qrdtprTtivs9LtdbTaODHDaCUClJKhZT9DVwAbKEWx+42A16UUuMxdVrewCyt9bOujch5lFKzgZGYGbVSgCeA74HvgAjgAPAvrfXJDY71llLqbGAlsJljdZaPYeqpG/Jx98U0HHljCkbfaa3/p5RqRgM+7uPZqz4e0Fpf6AnHrZTqhClFg6le/lpr/Wxtjt1tErUQQojyuUvVhxBCiApIohZCCDcniVoIIdycJGohhHBzkqiFEMLNSaIWQgg3J4laCCHcnCRqIYRwc/8Pe7Zc3ISOWfwAAAAASUVORK5CYII=\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": [
    "# plot the loss\n",
    "plt.plot(r.history['loss'], label='train loss')\n",
    "plt.plot(r.history['val_loss'], label='val loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('LossVal_loss')\n",
    "\n",
    "# plot the accuracy\n",
    "plt.plot(r.history['accuracy'], label='train acc')\n",
    "plt.plot(r.history['val_accuracy'], label='val acc')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('AccVal_acc')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20c6f7b0",
   "metadata": {},
   "source": [
    "#### VGG16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "4848b906",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 300 images belonging to 5 classes.\n",
      "Found 100 images belonging to 5 classes.\n",
      "Epoch 1/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 2.6209 - accuracy: 0.2433\n",
      "Epoch 00001: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 6s 507ms/step - loss: 2.6209 - accuracy: 0.2433 - val_loss: 2.1258 - val_accuracy: 0.2100\n",
      "Epoch 2/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.5079 - accuracy: 0.3833\n",
      "Epoch 00002: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 466ms/step - loss: 1.5079 - accuracy: 0.3833 - val_loss: 1.8785 - val_accuracy: 0.3500\n",
      "Epoch 3/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.1804 - accuracy: 0.5700\n",
      "Epoch 00003: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 498ms/step - loss: 1.1804 - accuracy: 0.5700 - val_loss: 1.3810 - val_accuracy: 0.4700\n",
      "Epoch 4/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.7039 - accuracy: 0.7233\n",
      "Epoch 00004: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 458ms/step - loss: 0.7039 - accuracy: 0.7233 - val_loss: 1.2298 - val_accuracy: 0.4900\n",
      "Epoch 5/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.5395 - accuracy: 0.8267\n",
      "Epoch 00005: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 487ms/step - loss: 0.5395 - accuracy: 0.8267 - val_loss: 1.5063 - val_accuracy: 0.4500\n",
      "Epoch 6/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.4851 - accuracy: 0.8433\n",
      "Epoch 00006: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 480ms/step - loss: 0.4851 - accuracy: 0.8433 - val_loss: 1.2826 - val_accuracy: 0.4900\n",
      "Epoch 7/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3957 - accuracy: 0.8667\n",
      "Epoch 00007: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 450ms/step - loss: 0.3957 - accuracy: 0.8667 - val_loss: 1.1634 - val_accuracy: 0.5600\n",
      "Epoch 8/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2918 - accuracy: 0.9333\n",
      "Epoch 00008: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 501ms/step - loss: 0.2918 - accuracy: 0.9333 - val_loss: 1.1207 - val_accuracy: 0.5800\n",
      "Epoch 9/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2409 - accuracy: 0.9533\n",
      "Epoch 00009: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 462ms/step - loss: 0.2409 - accuracy: 0.9533 - val_loss: 1.0813 - val_accuracy: 0.5300\n",
      "Epoch 10/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2117 - accuracy: 0.9667\n",
      "Epoch 00010: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 461ms/step - loss: 0.2117 - accuracy: 0.9667 - val_loss: 1.0818 - val_accuracy: 0.5400\n",
      "Epoch 11/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1600 - accuracy: 0.9800\n",
      "Epoch 00011: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 508ms/step - loss: 0.1600 - accuracy: 0.9800 - val_loss: 1.1751 - val_accuracy: 0.5400\n",
      "Epoch 12/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1517 - accuracy: 0.9900\n",
      "Epoch 00012: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.1517 - accuracy: 0.9900 - val_loss: 1.0432 - val_accuracy: 0.5700\n",
      "Epoch 13/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1298 - accuracy: 0.9933\n",
      "Epoch 00013: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 506ms/step - loss: 0.1298 - accuracy: 0.9933 - val_loss: 1.0573 - val_accuracy: 0.5800\n",
      "Epoch 14/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1200 - accuracy: 0.9967\n",
      "Epoch 00014: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 501ms/step - loss: 0.1200 - accuracy: 0.9967 - val_loss: 1.0819 - val_accuracy: 0.5600\n",
      "Epoch 15/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1045 - accuracy: 0.9967\n",
      "Epoch 00015: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 451ms/step - loss: 0.1045 - accuracy: 0.9967 - val_loss: 1.0812 - val_accuracy: 0.5700\n",
      "Epoch 16/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0896 - accuracy: 0.9967\n",
      "Epoch 00016: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 440ms/step - loss: 0.0896 - accuracy: 0.9967 - val_loss: 1.0606 - val_accuracy: 0.5600\n",
      "Epoch 17/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0869 - accuracy: 0.9967\n",
      "Epoch 00017: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 458ms/step - loss: 0.0869 - accuracy: 0.9967 - val_loss: 1.0325 - val_accuracy: 0.6000\n",
      "Epoch 18/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0881 - accuracy: 0.9933\n",
      "Epoch 00018: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 463ms/step - loss: 0.0881 - accuracy: 0.9933 - val_loss: 1.0341 - val_accuracy: 0.6100\n",
      "Epoch 19/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0665 - accuracy: 1.0000\n",
      "Epoch 00019: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 450ms/step - loss: 0.0665 - accuracy: 1.0000 - val_loss: 0.9829 - val_accuracy: 0.6200\n",
      "Epoch 20/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0629 - accuracy: 1.0000\n",
      "Epoch 00020: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 473ms/step - loss: 0.0629 - accuracy: 1.0000 - val_loss: 1.0107 - val_accuracy: 0.6400\n",
      "Epoch 21/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0544 - accuracy: 1.0000\n",
      "Epoch 00021: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 458ms/step - loss: 0.0544 - accuracy: 1.0000 - val_loss: 1.0173 - val_accuracy: 0.6300\n",
      "Epoch 22/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0626 - accuracy: 1.0000\n",
      "Epoch 00022: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 475ms/step - loss: 0.0626 - accuracy: 1.0000 - val_loss: 1.0361 - val_accuracy: 0.6100\n",
      "Epoch 23/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0470 - accuracy: 1.0000\n",
      "Epoch 00023: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 474ms/step - loss: 0.0470 - accuracy: 1.0000 - val_loss: 1.0373 - val_accuracy: 0.5800\n",
      "Epoch 24/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0533 - accuracy: 1.0000\n",
      "Epoch 00024: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 476ms/step - loss: 0.0533 - accuracy: 1.0000 - val_loss: 1.0462 - val_accuracy: 0.6000\n",
      "Epoch 25/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0485 - accuracy: 0.9967\n",
      "Epoch 00025: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 499ms/step - loss: 0.0485 - accuracy: 0.9967 - val_loss: 1.0313 - val_accuracy: 0.6200\n",
      "Epoch 26/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0475 - accuracy: 1.0000\n",
      "Epoch 00026: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 486ms/step - loss: 0.0475 - accuracy: 1.0000 - val_loss: 1.0235 - val_accuracy: 0.6000\n",
      "Epoch 27/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0379 - accuracy: 1.0000\n",
      "Epoch 00027: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 456ms/step - loss: 0.0379 - accuracy: 1.0000 - val_loss: 1.0311 - val_accuracy: 0.6100\n",
      "Epoch 28/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0386 - accuracy: 1.0000\n",
      "Epoch 00028: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 456ms/step - loss: 0.0386 - accuracy: 1.0000 - val_loss: 1.0447 - val_accuracy: 0.6000\n",
      "Epoch 29/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0428 - accuracy: 1.0000\n",
      "Epoch 00029: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 461ms/step - loss: 0.0428 - accuracy: 1.0000 - val_loss: 1.0381 - val_accuracy: 0.6200\n",
      "Epoch 30/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0350 - accuracy: 1.0000\n",
      "Epoch 00030: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 448ms/step - loss: 0.0350 - accuracy: 1.0000 - val_loss: 1.0482 - val_accuracy: 0.6200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 31/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0343 - accuracy: 1.0000\n",
      "Epoch 00031: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 474ms/step - loss: 0.0343 - accuracy: 1.0000 - val_loss: 1.0389 - val_accuracy: 0.6400\n",
      "Epoch 32/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0326 - accuracy: 1.0000\n",
      "Epoch 00032: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 475ms/step - loss: 0.0326 - accuracy: 1.0000 - val_loss: 1.0229 - val_accuracy: 0.6300\n",
      "Epoch 33/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0263 - accuracy: 1.0000\n",
      "Epoch 00033: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.0263 - accuracy: 1.0000 - val_loss: 1.0287 - val_accuracy: 0.6200\n",
      "Epoch 34/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0281 - accuracy: 1.0000\n",
      "Epoch 00034: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 457ms/step - loss: 0.0281 - accuracy: 1.0000 - val_loss: 1.0293 - val_accuracy: 0.6400\n",
      "Epoch 35/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0272 - accuracy: 1.0000\n",
      "Epoch 00035: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 495ms/step - loss: 0.0272 - accuracy: 1.0000 - val_loss: 1.0489 - val_accuracy: 0.6300\n",
      "Epoch 36/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0291 - accuracy: 1.0000\n",
      "Epoch 00036: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 525ms/step - loss: 0.0291 - accuracy: 1.0000 - val_loss: 1.0314 - val_accuracy: 0.6200\n",
      "Epoch 37/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0242 - accuracy: 1.0000\n",
      "Epoch 00037: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 477ms/step - loss: 0.0242 - accuracy: 1.0000 - val_loss: 1.0366 - val_accuracy: 0.6400\n",
      "Epoch 38/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0249 - accuracy: 1.0000\n",
      "Epoch 00038: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 471ms/step - loss: 0.0249 - accuracy: 1.0000 - val_loss: 1.0628 - val_accuracy: 0.6000\n",
      "Epoch 39/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0268 - accuracy: 1.0000\n",
      "Epoch 00039: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 510ms/step - loss: 0.0268 - accuracy: 1.0000 - val_loss: 1.0559 - val_accuracy: 0.6300\n",
      "Epoch 40/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0237 - accuracy: 1.0000\n",
      "Epoch 00040: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 497ms/step - loss: 0.0237 - accuracy: 1.0000 - val_loss: 1.0631 - val_accuracy: 0.6200\n",
      "Epoch 41/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0224 - accuracy: 1.0000\n",
      "Epoch 00041: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 444ms/step - loss: 0.0224 - accuracy: 1.0000 - val_loss: 1.0594 - val_accuracy: 0.6400\n",
      "Epoch 42/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0243 - accuracy: 1.0000\n",
      "Epoch 00042: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 467ms/step - loss: 0.0243 - accuracy: 1.0000 - val_loss: 1.0431 - val_accuracy: 0.6500\n",
      "Epoch 43/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0213 - accuracy: 1.0000\n",
      "Epoch 00043: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 444ms/step - loss: 0.0213 - accuracy: 1.0000 - val_loss: 1.0412 - val_accuracy: 0.6300\n",
      "Epoch 44/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0215 - accuracy: 1.0000\n",
      "Epoch 00044: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 470ms/step - loss: 0.0215 - accuracy: 1.0000 - val_loss: 1.0561 - val_accuracy: 0.6400\n",
      "Epoch 45/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0180 - accuracy: 1.0000\n",
      "Epoch 00045: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 478ms/step - loss: 0.0180 - accuracy: 1.0000 - val_loss: 1.0529 - val_accuracy: 0.6400\n",
      "Epoch 46/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0179 - accuracy: 1.0000\n",
      "Epoch 00046: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 505ms/step - loss: 0.0179 - accuracy: 1.0000 - val_loss: 1.0694 - val_accuracy: 0.6400\n",
      "Epoch 47/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0152 - accuracy: 1.0000\n",
      "Epoch 00047: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 508ms/step - loss: 0.0152 - accuracy: 1.0000 - val_loss: 1.0600 - val_accuracy: 0.6400\n",
      "Epoch 48/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0174 - accuracy: 1.0000\n",
      "Epoch 00048: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 15s 2s/step - loss: 0.0174 - accuracy: 1.0000 - val_loss: 1.0652 - val_accuracy: 0.6400\n",
      "Epoch 49/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0174 - accuracy: 1.0000\n",
      "Epoch 00049: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 6s 546ms/step - loss: 0.0174 - accuracy: 1.0000 - val_loss: 1.0797 - val_accuracy: 0.6400\n",
      "Epoch 50/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0172 - accuracy: 1.0000\n",
      "Epoch 00050: saving model to best_vgg16_model.h5\n",
      "10/10 [==============================] - 5s 481ms/step - loss: 0.0172 - accuracy: 1.0000 - val_loss: 1.0600 - val_accuracy: 0.6500\n"
     ]
    }
   ],
   "source": [
    "r, model, test_set = create_model(VGG16, 'best_vgg16_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "41e6e65d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 1s - loss: 1.0069 - accuracy: 0.6500 - 1s/epoch - 264ms/step\n",
      "Restored model, accuracy: 65.00%\n"
     ]
    }
   ],
   "source": [
    "# VGG16 Model\n",
    "model.save('model_vgg16.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "85844156",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 2s - loss: 1.0069 - accuracy: 0.6500 - 2s/epoch - 379ms/step\n",
      "Restored model, accuracy: 65.00%\n"
     ]
    }
   ],
   "source": [
    "# VGG!6 model while checkpointing\n",
    "model=load_model('best_vgg16_model.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "a64748cd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAsTElEQVR4nO3deXxU9b3/8ddntuwhOwmBkCAouwEiplJZ1KsgbbXV9mK1Lr23/ry1vbWLP73dW29/9Vq74VKvbW217rXa2rovIGjdAEFBQFklLFmAhISsM/P9/fE9WQgJmSSTTGbm83w85jHbmZnPyfI+3/me7/keMcaglFIq+rkiXYBSSqnw0EBXSqkYoYGulFIxQgNdKaVihAa6UkrFCE+kPjgnJ8cUFxdH6uOVUioqrV27tsYYk9vTcxEL9OLiYtasWROpj1dKqagkIrt7e67PLhcRSRSRt0Rkg4hsEpEf9bCMiMhyEdkmIu+KyOzBFq2UUqp/QmmhtwBnGWMaRMQLvCoizxhj3uiyzBJgknM5HfiNc62UUmqY9NlCN1aDc9frXLofXnoBcJ+z7BtAhogUhLdUpZRSJxJSH7qIuIG1wETgDmPMm90WKQT2dLlf4Ty2v9v7XA1cDVBUVDTAkpVSI11bWxsVFRU0NzdHupSolZiYyNixY/F6vSG/JqRAN8YEgFIRyQCeEJHpxpiNXRaRnl7Ww/vcDdwNUFZWppPIKBWjKioqSEtLo7i4GJGe4kGdiDGGgwcPUlFRQUlJSciv69c4dGNMLbASWNztqQpgXJf7Y4F9/XlvpVTsaG5uJjs7W8N8gESE7Ozsfn/DCWWUS67TMkdEkoBzgC3dFnsSuNwZ7VIO1Blj9qOUilsa5oMzkJ9fKC30AmCFiLwLvA28YIz5h4hcIyLXOMs8DewAtgG/Bb7c70pCtPVAPbc+t5XDR1uH6iOUUioqhTLK5V1jzCxjzExjzHRjzI+dx+8yxtzl3DbGmGuNMScZY2YYY4bsiKGdNUe5fcU29tY2DdVHKKWiXG1tLXfeeeeAXnv++edTW1sb8vI//OEPufXWWwf0WeEWdXO5ZKX4ADjcqC10pVTPThTogUDghK99+umnycjIGIKqhl4UBrodwnNIu1yUUr248cYb2b59O6WlpVx//fWsXLmSRYsW8fnPf54ZM2YAcOGFFzJnzhymTZvG3Xff3fHa4uJiampq2LVrF1OmTOFLX/oS06ZN49xzz6Wp6cQ9A+vXr6e8vJyZM2fy6U9/msOHDwOwfPlypk6dysyZM1m2bBkAr7zyCqWlpZSWljJr1izq6+sHvd4Rm8tloDKTbQu9trEtwpUopULxo79v4v19R8L6nlPHpPODT07r9fmbb76ZjRs3sn79egBWrlzJW2+9xcaNGzuGAd5zzz1kZWXR1NTEaaedxkUXXUR2dvYx7/Phhx/y0EMP8dvf/pbPfe5z/OUvf+Gyyy7r9XMvv/xybrvtNhYsWMD3v/99fvSjH/GrX/2Km2++mZ07d5KQkNDRnXPrrbdyxx13MG/ePBoaGkhMTBzcD4UobKGPSvIioi10pVT/zJ0795gx3cuXL+fUU0+lvLycPXv28OGHHx73mpKSEkpLSwGYM2cOu3bt6vX96+rqqK2tZcGCBQBcccUVrFq1CoCZM2dy6aWXcv/99+Px2Hb0vHnz+MY3vsHy5cupra3teHwwoq6F7nG7GJXk1T50paLEiVrSwyklJaXj9sqVK3nxxRd5/fXXSU5OZuHChT2O+U5ISOi47Xa7++xy6c1TTz3FqlWrePLJJ7npppvYtGkTN954I0uXLuXpp5+mvLycF198kcmTJw/o/dtFXQsdICvZpy10pVSv0tLSTtgnXVdXR2ZmJsnJyWzZsoU33nij12VDNWrUKDIzM1m9ejUAf/rTn1iwYAHBYJA9e/awaNEibrnlFmpra2loaGD79u3MmDGDG264gbKyMrZs6X54T/9FXQsdIDPFpy10pVSvsrOzmTdvHtOnT2fJkiUsXbr0mOcXL17MXXfdxcyZMznllFMoLy8Py+fee++9XHPNNTQ2NjJhwgT+8Ic/EAgEuOyyy6irq8MYw9e//nUyMjL43ve+x4oVK3C73UydOpUlS5YM+vPFmMhMqVJWVmYGeoKLf7/3bfbWNvPM184Mc1VKqXDYvHkzU6ZMiXQZUa+nn6OIrDXGlPW0fFR2uWQm+/RIUaWU6iYqAz0rxcehxlYi9e1CKaVGoqgM9MwUH63+II2tJz7iSyml4klUBnqWc3CRjnRRSqlOURnomSl6tKhSSnUXlYHeMZ+LDl1USqkOURno7fO56EgXpVS4pKam9uvxkSgqA719Cl3tQ1dKqU5RGejpiV5conOiK6V6dsMNNxwzH/oPf/hDfv7zn9PQ0MDZZ5/N7NmzmTFjBn/7299Cfk9jDNdffz3Tp09nxowZPPLIIwDs37+f+fPnU1payvTp01m9ejWBQIArr7yyY9lf/vKXYV/HnkTlof8ul5Cp87koFR2euREOvBfe98yfAUtu7vXpZcuWcd111/HlL9uzYT766KM8++yzJCYm8sQTT5Cenk5NTQ3l5eV86lOfCun8nY8//jjr169nw4YN1NTUcNpppzF//nwefPBBzjvvPL7zne8QCARobGxk/fr17N27l40bNwL06wxIgxGVgQ6QkawzLiqlejZr1iyqqqrYt28f1dXVZGZmUlRURFtbG9/+9rdZtWoVLpeLvXv3UllZSX5+fp/v+eqrr3LJJZfgdrsZPXo0CxYs4O233+a0007ji1/8Im1tbVx44YWUlpYyYcIEduzYwVe/+lWWLl3KueeeOwxrHcWBnpWiLXSlosIJWtJD6eKLL+axxx7jwIEDHWcJeuCBB6iurmbt2rV4vV6Ki4t7nDa3J70dmT5//nxWrVrFU089xRe+8AWuv/56Lr/8cjZs2MBzzz3HHXfcwaOPPso999wTtnXrTVT2oUP7fC46Dl0p1bNly5bx8MMP89hjj3HxxRcDdtrcvLw8vF4vK1asYPfu3SG/3/z583nkkUcIBAJUV1ezatUq5s6dy+7du8nLy+NLX/oS//Zv/8a6deuoqakhGAxy0UUXcdNNN7Fu3bqhWs1jRHUL/Z09tZEuQyk1Qk2bNo36+noKCwspKCgA4NJLL+WTn/wkZWVllJaW9uuEEp/+9Kd5/fXXOfXUUxERbrnlFvLz87n33nv52c9+htfrJTU1lfvuu4+9e/dy1VVXEQwGAfjpT386JOvYXVROnwvwP89u4Xerd/DBfy8JaYeGUmr46PS54REX0+eCnc+lLWBoaPFHuhSllBoRojbQ2+dz0X50pZSy+gx0ERknIitEZLOIbBKRr/WwzEIRqROR9c7l+0NTbiedz0WpkU3PVzA4A/n5hbJT1A980xizTkTSgLUi8oIx5v1uy602xnyi3xUMkM7notTIlZiYyMGDB8nOztZ9XANgjOHgwYMkJib263V9BroxZj+w37ldLyKbgUKge6APK53PRamRa+zYsVRUVFBdXR3pUqJWYmIiY8eO7ddr+jVsUUSKgVnAmz08/TER2QDsA75ljNnUw+uvBq4GKCoq6leh3WW0t9C1y0WpEcfr9VJSUhLpMuJOyDtFRSQV+AtwnTHmSLen1wHjjTGnArcBf+3pPYwxdxtjyowxZbm5uQMs2UpP9OB2ibbQlVLKEVKgi4gXG+YPGGMe7/68MeaIMabBuf004BWRnLBWenxN9mhRbaErpRQQ2igXAX4PbDbG/KKXZfKd5RCRuc77HgxnoT3JSvFqC10ppRyh9KHPA74AvCci653Hvg0UARhj7gIuBv5DRPxAE7DMDMOYJZ3PRSmlOoUyyuVV4ITjjowxtwO3h6uoUGWl+NhW1TDcH6uUUiNS1B4pCvZoUe1DV0opK6oDPSvZx+HGNoJBPSJNKaWiOtAzU3wEgob6Zp2gSymlojrQdT4XpZTqFNWB3n60qA5dVEqpKA/0LJ2gSymlOkR3oLdP0KVdLkopFd2B3nmSCw10pZSK6kBP8bnxuV3aQldKKaI80EWEzBQvtXr4v1JKRXegg53PRVvoSikVA4GeleLTPnSllCIGAj0zRVvoSikFMRDoWcnaQldKKYiBQM9M8VHb1EZAJ+hSSsW56A/0ZC/GQF2TjnRRSsW3qA/0jqNFtdtFKRXnoj7QM9vnc9Edo0qpOBedgV5fCc4pS7WFrpRSVvQF+oZH4Ocnw6EdgM7nopRS7aIv0EdPs9d71wFdptBt1J2iSqn4Fn2BnjsZvMmwdy0AST43iV6X9qErpeJe9AW62wMFpR2BDraVrn3oSql4F32BDlA4G/ZvgIDtZsnU+VyUUipaA30OBFqgchNgR7rofC5KqXjXZ6CLyDgRWSEim0Vkk4h8rYdlRESWi8g2EXlXRGYPTbmOwjn22ul2ydD5XJRSKqQWuh/4pjFmClAOXCsiU7stswSY5FyuBn4T1iq7yyiC5JyOQM9K9mofulIq7vUZ6MaY/caYdc7temAzUNhtsQuA+4z1BpAhIgVhr7adiG2lO4GemeLjSLOftkBwyD5SKaVGun71oYtIMTALeLPbU4XAni73Kzg+9BGRq0VkjYisqa6u7mep3T9xDlRvheYjHUeL1upYdKVUHAs50EUkFfgLcJ0x5kj3p3t4yXHz2Rpj7jbGlBljynJzc/tXaXeFc+xH7F+v87kopRQhBrqIeLFh/oAx5vEeFqkAxnW5PxbYN/jyTqDQ2e+6d21HC113jCql4lkoo1wE+D2w2Rjzi14WexK43BntUg7UGWP2h7HO4yVnQdYE2LtWW+hKKQV4QlhmHvAF4D0RWe889m2gCMAYcxfwNHA+sA1oBK4Ke6U9KZwDu//ZZcZF7UNXSsWvPgPdGPMqPfeRd13GANeGq6iQFc6B9/5MRqAG0Ba6Uiq+ReeRou2cA4wSqzaQ4nPrWHSlVFyL7kDPnwEuD+xdq0eLKqXiXnQHujfJzo/ujHTR+VyUUvEsugMdnCNG3yEr2aMtdKVUXIuNQG+p4xRPpbbQlVJxLTYCHZhiPuSwDltUSsWx6A/0nJPBl8qElq00tPhp9esEXUqp+BT9ge5yw5hZFDa+D0CtdrsopeJU9Ac6QOFsMuu34qNN+9GVUnErRgJ9Du5gG1Nktx5cpJSKWzET6ACnurbrjlGlVNyKjUBPLySYksepru3sq22KdDVKKRURsRHoIrjGljHbvYOdB49GuhqllIqI2Ah0gMLZlLCPysrKSFeilFIREUOBbvvR51c9AAe3R7gYpZQafqGc4CI6jJ1LTcpErjj6ONz2OGSWwMRz7KXkTPClRLpCpZQaUrHTQk9I5a3F/2B+yy/Zd8ZNkHsKrH8AHvpX+J9iWP9QpCtUSqkhFTuBDpTkpPCRGc3a0RfD5x+BG3bB5X+zrfW37o50eUopNaRiKtCLs223ys4aZ6SLJwEmLIRpn4Z970DjocgVp5RSQyymAj3J52bMqMTOQG930iLAwM5VEalLKaWGQ0wFOkBJbgo7ugd64RzwpcGOFZEpSimlhkHMBfqEnFR2VjdgjOl80O21I122vwxdH1dKqRgSc4FekpPCkWb/8ZN0TVgEtR/BoR2RKUwppYZY7AV6rt0xely3y0mL7LV2uyilYlTMBfqEHGekS3W3QM+eCOljYbsGulIqNvUZ6CJyj4hUicjGXp5fKCJ1IrLeuXw//GWGrjAjCa9bjm+hi8BJC2Hnagj4I1KbUkoNpVBa6H8EFvexzGpjTKlz+fHgyxo4j9tFUVYyO2sajn/ypLOgpc6OSVdKqRjTZ6AbY1YBUXVEzoTc1OPHogOULARE+9GVUjEpXH3oHxORDSLyjIhM620hEblaRNaIyJrq6uowffTxJuSksOtgI4FgtyGKKdlQMFP70ZVSMSkcgb4OGG+MORW4DfhrbwsaY+42xpQZY8pyc3PD8NE9K8lJodUf7PnsRRMWQcVb0FI/ZJ+vlFKRMOhAN8YcMcY0OLefBrwikjPoygahJKeXoYtghy8G/bDrtWGuSimlhtagA11E8kVEnNtznfc8ONj3HYz2seg7q3vYMTquHDyJ2o+ulIo5fZ7gQkQeAhYCOSJSAfwA8AIYY+4CLgb+Q0T8QBOwzJjIHl+fm5pAaoKn5x2j3kQYf4b2oyulYk6fgW6MuaSP528Hbg9bRWEgIkzoaZKudhMWwQvfg7q9MKpweItTSqkhEnNHirYryUnpuYUOXaYBWDls9Sil1FCL6UDfW9tEc1vg+CfzpkFKrvajK6ViSkwHujGw+2Dj8U+6XPZMRjtWQjA43KUppdSQiNlAn5CTCtDzFABg+9GPVkPVpmGsSimlhk7MBnpxTjLQy1h06OxH19EuSqkYEbOBnpboJS8t4fhpdNulj4GcU2D9A1D5/vAWp5RSQyBmAx36GOkCcNZ3oX4/3DUPnvxPqD8wfMUppVSYxXSgT8jtI9Cnfgr+cz2cfg2sfxCWz4aV/wOtJ3iNUkqNUDEd6CU5KRw82kpdY1vvCyVnweKfwrVvwqRzYOX/s8G+4ZHhK1QppcIgxgPdjnTZ0dtIl66yT4LP3QdffB4yxsETV8O2l4a4QqWUCp8YD3Rnkq4Tdbt0V3Q6XPEPu8P0ya9Cc90QVaeUUuEV04FelJWM2yX9C3SwE3hd+Bu7w/S57wxNcUopFWYxHeg+j4txmUm9j0U/kbFzYN7X4J0/wYcvhL84pZQKs5gOdHCGLvY2Fr0vC/8LcifbIY1NtWGtSymlwi0OAt2eMHpAU7R7EuDCO6GhUrtelFIjXuwHem4KTW0BDhxpHtgbFM6Bj18H6++HD54La21KKRVOMR/oE9pHugy02wVgwQ2QNxX+/jVoOhymypRSKrxiP9BzT3DC6FB1dL1UwbP/FabKlFIqvGI+0EenJZLkdbOtKoSDi05kzCw48xuw4SG7k7T5SHgKVEqpMOnznKLRzuUSZo/P4LVtNYN/swU3gL8FXr8dtr8Mn7qtcxpepZSKsJhvoQMsOiWPD6sa2HOoh7MX9YfbC+feBF98znbD/OlC+MfXoaU+LHUqpdRgxEWgnzU5D4AVW6vC84bj5sI1r8LHvgJr/gB3ngE7Xgn99cbAvnd0HnalVFjFfJcLwITcVIqzk3l5SxWXf6w4PG/qTYLzfgJTPgl//TLc9ykoKIXij8P4eTD+Y5CU2bm8vwV2roatT8HWZ+y0Agic+U17AJM7Ln4VSqkhFDcpsmhyHg+++RFNrQGSfO7wvXFRuW2tv/kb2PYyvPVb28eOwOjpMP4MaDhgZ25sbQBvCkw8C045H3a/BqtvhV2vwkW/s7M8KqXUAMmAjqAMg7KyMrNmzZph+7zVH1bzhd+/xe+vKOPsKaOH7oPammHvWhvWu1+DPW9BQjqcssSGeMl8O/lXu3f/bPvhXS644A7b4g+Xowch0ArpBeF7T6VURInIWmNMWU/P9dlCF5F7gE8AVcaY6T08L8CvgfOBRuBKY8y6wZUcfnNLskj2uXl5S9XQBro3EYrn2QtAMADiApGel5/5WTsR2GNfhEcug9P+Hc79b9ulM1D+Vnj9NnjlZ+Bvgrxp9uQdE8+BceXg8Q38vZVSI1YoXS5/BG4H7uvl+SXAJOdyOvAb53pESfC4+fjEHFZsqcIYg/QWsOHmCqF7J2uCPbHGSz/qHBJZsgDyZ0D+TBg9FXwpoX3ejlfg6W9BzQe2tV9YBttehNfvhNd+Db40mLDAfmOYeiEkpPb9noE22L/BTlTW1/IBP2x8zH6WN8kO7Rw9LbTalYp1bc1w4F27fy1nUtjfPqQuFxEpBv7RSwv9f4GVxpiHnPtbgYXGmP0nes/h7nIBePitj7jx8fd49rozmZyfPqyfHbIPX4TXfmV/6R0n1xDInmgDvmCmE/SnQmpu5+vqD8Dz34X3/gyZxXD+rTDpXzqfb6mHnavsVMDbXoS6PeBLhRmfhTlXwpjSY+swxob4hoftezbW2P7/aRdC6eeh6AzbTdQu0AbvPmr3CRzaYb8VHK2G5lo4+/tQfu2xyyvVX/5WO5igtQFaGqC13rl2DhocPc1O0eFJ6P09ggE4uB2q3rcNjrQCSB8Dydm9f4tubbRTfgTbIHGU7UINpaFmjP1fqFgDe9dAxdtwYKN9n499xQ6qGIBBdbmEoBDY0+V+hfPYcYEuIlcDVwMUFRWF4aP7Z5EzfPHlLVUjN9AnnWMvxtjQPfCevex/1/5BbHq8c9m0AtuCzxxvg9ffAgtutJOJde+ySUiDyUvtxRjbt7/2j/bI17V/sCN0yq6C4jNhy1P28ar3we3r7P/f/RpsfALWPwAZ422wT7/Y2bn7c6jdbev51wfs8k2H7Pw3z3/XTmx24W90x+9I0nwEqrfYMMsYP/JGWjUesn/zH70Be960+6b8fUyy5/JA7hTb8Ck41Z5a8tBOqNxo/48q37fdkN25EyAt34a7uGyAt1+O+0yxoZ44CpJG2YZRoNVucAIt9v8w0Nq50QHbGCqcDWd8xX5rHjc3LD+i7sLRQn8K+Kkx5lXn/kvA/zXGrD3Re0aihQ6wdPlqkn1u/nzNGcP+2WHReKgz5A+8a4O+5gN7xOqSW+wfcH80HbY7Ztf+wQZ4u8IyKL0Epn3Gnki7XWsjbPkHrH8QdqwEnL+fMbPtkbQnn3dsS8cYuwF45gYQNyy91X4rELHdM0f22g3B4V1Qt9d+k2h1Wl2tR+0/RVujrSF9DKQX2uu0MXZnb6DVfjup3w9H9tvr9iGh2RPtzyN7ov16mzbGfktoOgxVm6Fyk13nqs1Q+xHknGxn1yycA2PLIDWvcz38LTYU9q6zxxDse8ceaNb+bSl/hm0hJg6goeBvsT+boQzUgB+qN3dpLa61Yd7++3N5IavE+Zk5F18KmKBt1Qb9YAL2NtgWqrjttctjQzDot/MdHa2Chmo77fTRansuAV9yZ+s2Md1eJ6Q5tbXZ32OgzbZe/S22tuotzmd5bDiPK4e8KfZ1Cam2+zAh1dYZ8EOl0/DZv8FeGrscHZ6Y4fyunEveVPt59fvgiHNp/xsCSMqw3SLtl+QsW0fzEfuts7nOrldznf1bdfvsN4Ou195k211aWGbrDqVVH4ITtdDjqssF4OfPb+WOFdtY971/ISM5RnYOBgOD/2Mxxv6zV7wFk84NrX+vrgI2/x2yJ8HEs3v/ygo2sB//P7DnDduKbzliXx/0d1lIbGvHl2IvCan2vifRtvaP7LOBQW9/s2JDOK3A/kwObbcbg3aeJBsmDZWdjyWOst1Do8baAKncZIMLYNQ4W2v9fhvmgVb7eHKObW0F2uxGtfFg5/tlTbBhkTXBBmTWBHtJL7S/o/pKZ0O8oXODfHinfa3LY2v0Jtmd654kW0vX1l/ACTy3zwmaLqGTlGXDpPmIDZqul6PVnS3TpCy7wSoss+HWdAgOboOaD213xKEd9rMGyp0AqaNtl2DqaBumbUdtXS3ttR2xG28Ru2F0++zF5bH3M0vs+X3HldsNrC+5fzUYY39vB7fb30N64Yn/PqPIUAf6UuAr2FEupwPLjTF9fp+IVKCv++gwn7nzn/x6WSkXlBYO++fHtWAA/nkbfPCs/QfLHG/7+zOc6/TCvlup/lY7rv/IPtu6dyc4/aAFkJJ37Os7/qm3OZfttlWVe7IN8bwpzlfsLv/orY02aPeutRu4A+/a9y+cbb+FFM62Qd/+mvbPaO8WO/Cu3TAc3tW5AQAbVr5UG57tMovtBiNvqm3h+pvsTrOu1+J2Wnxeu67tt/2tx3YLNB1yugda7UYrcVTnJSEdUnJsK7dwjt3AnCjcggG7sW1rsgHrcjmtcHdnwyEYcFrsfggG7W1x2xBPSI+Z8ByJBhXoIvIQsBDIASqBHwBeAGPMXc6wxduBxdhhi1cZY/pM6kgFeiBoOO0nL3LmpBx+vWzWsH++ihPBgN3oHN5pW7yHdtrAzZtiQzx/ug1bpfppUDtFjTGX9PG8Aa4dYG3Dzu0SFp6cy8tbqwgEDW6XtiTUEHC57Q7gjHH2YDKlhkFcjiNbNDmP2sY23vlIzz6klIodcRno80/Oxe0SXt4SptkXlVJqBIjLQB+V5GXO+EwNdKVUTInLQAc7R/qWA/Xsq+3hIAOllIpCcR3oEMaTXiilVITFbaBPykulMCOJlzdroCulYkPcBrqIsHRmAa98UE3VkT7mh1BKqSgQt4EOcMncIvxBw8Nv7+l7YaWUGuHiOtBLclI4c1IOD775Ef5AMNLlKKXUoMR1oANcVj6eA0eaeUmHMCqlolzcB/rZk/MoGJXI/W/sjnQpSik1KHEf6B63i0vmFrH6wxp21hyNdDlKKTVgcR/oAMtOG4fHJTygrXSlVBTTQAfy0hM5b1o+f15bQXNbINLlKKXUgGigOy4rH09dUxt/37Av0qUopdSAaKA7yidkMTEvlfvf/CjSpSil1IBooDtEhMtOL2LDnlreq6iLdDlKKdVvGuhdfGbOWJK8bh3CqJSKShroXaQnerlw1hj+tmEvdY1tkS5HKaX6RQO9m0tPH09zW5C/rKuIdClKKdUvGujdTC8cxayiDO5/Yzf2/NdKKRUdNNB78IXy8eyoOcqr22oiXYpSSoVMA70HS2cWkJPq44+v7Yp0KUopFTIN9B4keNx8fm4RL2+tYvdBnd9FKRUdNNB7cWn5eNwi3Pe6DmFUSkWHkAJdRBaLyFYR2SYiN/bw/EIRqROR9c7l++EvdXiNTk9kyYwCHl2zh6Mt/kiXo5RSfeoz0EXEDdwBLAGmApeIyNQeFl1tjCl1Lj8Oc50RceUZ46lv9vPEO3sjXYpSSvUplBb6XGCbMWaHMaYVeBi4YGjLGhlmF2UyvTCde/+5S4cwKqVGvFACvRDoehblCuex7j4mIhtE5BkRmdbTG4nI1SKyRkTWVFdXD6Dc4SUiXHlGCR9WNfDP7QcjXY5SSp1QKIEuPTzWvbm6DhhvjDkVuA34a09vZIy52xhTZowpy83N7VehkfKJmQVkpfj4gw5hVEqNcKEEegUwrsv9scAxk4YbY44YYxqc208DXhHJCVuVEZTotUMYX9pSyZ5DjZEuRymlehVKoL8NTBKREhHxAcuAJ7suICL5IiLO7bnO+8ZMH8Wl5UW4RLjv9V2RLkUppXrVZ6AbY/zAV4DngM3Ao8aYTSJyjYhc4yx2MbBRRDYAy4FlJob2IhaMSmLxtHweeXsPja06hFEpNTJ5QlnI6UZ5uttjd3W5fTtwe3hLG1munFfMU+/t54l39nLp6eMjXY5SSh1HjxQNUdn4TKYW6BBGpdTIpYEeIhHhynnFfFDZwC9e+EBDXSk14oTU5aKsi2aP5e2dh7jt5W3UNLRy0wXT8Lh1m6iUGhk00PvB7RJuuXgmeekJ3LFiOwcbWlh+ySwSve5Il6aUUtrl0l8iwvXnTeaHn5zKC5srufz3b+n5R5VSI4IG+gBdOa+E2y6ZxTt7DvO5/32dA3XNkS5JKRXnNNAH4RMzx3DvVXPZW9vEZ+58jTd3xMyxVEqpKKSBPkhnTMzh4avLCRr417vf4NLfvcHbuw5FuiylVBzSQA+D6YWjWPGthXx36RS2Hqjns3e9zmW/e5O1uzXYlVLDRyI1nrqsrMysWbMmIp89lJpaA9z/xm7+d9V2ahpaOXNSDl+cV8IZE7NJ8OhoGKXU4IjIWmNMWY/PaaAPjcZWPw+88RF3vbKdg0dbSU3wsGhyHudNG83CU/JITdARo0qp/tNAj6DmtgD/3F7DcxsreXFzJQePtuJzu5g3MZtzp+Vz1uQ8RqcnRrpMpVSUOFGgazNxiCV63Zw1eTRnTR5NIGhYu/swz206wHObDrBi63sAzCgcxVmT8zh7Sh7Tx4zC5erpnCJKKXVi2kKPEGMMH1Q28OLmSl7eUsW6jw5jDOSmJXDWKXmcM3U0H5+YQ5JP+92VUp20yyUKHGxo4ZUPqnlpSxWrtlZT3+InweNi3sQczpkymrOnaNeMUkoDPeq0+oO8tfMQL262/e4Vh5sAmDl2FOdNy2fx9HxOyk2NcJVKqUjQQI9iXbtmnn+/kg17agE4eXQqS6YXsGRGPqeMTsM5A6BSKsZpoMeQfbVNPLfpAM+8d4C3dx/CGCjJSaE4O5kEj5tEr4tEr5tEr5sEr4vc1AROyktlYm4qYzKScOsOV6WimgZ6jKqqb+b5TZW8tLmSmoZWWvwBmtuCNLcF7MUfpNUf7Fg+weOiJCeFk/JSmTYmnXOn5jMxT7tulIomGuhx7NDRVrZXN7C9qoHt1Q1sq2pgW3UDew7ZfvlJeaksmZ7PedPzmVqQrl03So1wGujqOPtqm3h+0wGe2XiAt3cdImigKCuZMyflIALNbUGa2gI0twZoagvQFggyLiuZKfnpnJKfxuSCNHJTE3QDoNQw00BXJ1TT0MIL71fy7MYDrNt9GK/HRZK3sz8+yevG5RJ21Rylqr6l43XZKT5OyU8jK8WH1+3C4xK8Hhdel+BxuwgaQ1NrgEbn0tTmp6k1gNsltusnN5WTclOZmJfK2MwkPZ2fUiHQQFdhc+hoK1sOHGHrgXq27K9na2U9R5rb8AcMbYEgbQGDPxikzR/EJUKSz02yz+6kTfa5SfZ5aPUH2VFzlJqGzo2Dz+2iOCeZyfnpTClIZ0pBGlML0slN028BSnWlh/6rsMlK8XHGSTmccVLOoN+rrrGN7TXt/ftH2VZVz9rdh3lyw75jPu+U0Wkk+dwEjSEQNBhDx+20RA+5aQnkpiaQ41znpiWQkuDB72xcAkGDP2iXb9/o2Gu707g1ECRoID89kbGZSYzLStbJ01RU0r9aFTGjkr3MLspkdlHmMY/XNbax+cARNu8/csy3ALdLcIngEnvCbkHYW9vMhoo6Dja0EAzjl83MZC9jM5MZm5lEeqLXdj/53CR63CT53CR6XCR43SR4XPg8LhI8bnweFz63C59HELG1CthrAREIBiHgbIy6bqASvS5SEjwk+9yk+Dwk+ex767cT1R8hBbqILAZ+DbiB3xljbu72vDjPnw80AlcaY9aFuVYVJ0YleymfkE35hOyQXxMIGg4dbaW6voXqhhYaW/x43C48bsHjEtwuweNy4XYJCR4XXrcLr1vwul0keFwYYH9dMxWHG9lzqMleH27ig8p6Glps33/3YaBDze0SW6PLhbuH9Wi/7+54vNvz7s7HXc4GxuUCwW5g2jc07d9UWvyd122BID6Pi9QET+cl0UNagocEr51fSKTzvaSj3s6frc/5ObtdQrDbt6T2b03ttXvdgttl98N4XELQ4CwT7Pim5Xc2fp6O1xz7cxBnY283np0bf6/b1bGxbf/d+5wNsdflwuuRjn1AQ70BNcbQ4vwNJXrDP09Tn4EuIm7gDuBfgArgbRF50hjzfpfFlgCTnMvpwG+ca6WGhdsltuslLWHA7zE6PZHScRknXCYQNB3j/Vv8AVrabJeNDcKAE4YGYzq7htqvgwYnwGyYul2C20nDlrYgR1v9dgdyi5+jrQEaW/12n0TAEHACzQaivR8w2McDtrXvD7Yvazrq7Nrd1FmHwUDHfa/bCTuvvU5L9OBzu2gNBGlo8XOgrpmGFj8NzX4aWv1EaLfbsPC67Tc/g/15GTp/hyLglmM3ol03mO4u13bjAK2BoP1b6XZcyH8sPIkbFk8Oe/2htNDnAtuMMTsARORh4AKga6BfANxn7B7WN0QkQ0QKjDH7w16xUhHkdgnJPg/JvkhXEhnBoKEtGOwIdRt6NvACxnTsHG91Wvnt+ys6vyk5rXC33ZgFDce02Ntb5OK0rN2uzm8oXuco5/YNlL0OdmzI2msJGtsSDjobrDZ/kJb2/SX+zv0m/kCQ1oDB374/xam1PbyFY7+FAB0by/bPb9+QBtsfN85tY39WPo+LRK/tkkvwukj02AECs4syhuT3E0qgFwJ7utyv4PjWd0/LFALHBLqIXA1cDVBUVNTfWpVSEeZyCQkundJ5pApl4G9PnUrdv3SFsgzGmLuNMWXGmLLc3NxQ6lNKKRWiUAK9AhjX5f5YYN8AllFKKTWEQgn0t4FJIlIiIj5gGfBkt2WeBC4Xqxyo0/5zpZQaXn32oRtj/CLyFeA57LDFe4wxm0TkGuf5u4CnsUMWt2GHLV41dCUrpZTqSUjj0I0xT2NDu+tjd3W5bYBrw1uaUkqp/tDZkJRSKkZooCulVIzQQFdKqRgRselzRaQa2D3Al+cANWEsJ5rE67rrescXXe/ejTfG9HggT8QCfTBEZE1v8wHHunhdd13v+KLrPTDa5aKUUjFCA10ppWJEtAb63ZEuIILidd11veOLrvcARGUfulJKqeNFawtdKaVUNxroSikVI6Iu0EVksYhsFZFtInJjpOsZKiJyj4hUicjGLo9licgLIvKhc515oveIRiIyTkRWiMhmEdkkIl9zHo/pdReRRBF5S0Q2OOv9I+fxmF7vdiLiFpF3ROQfzv2YX28R2SUi74nIehFZ4zw2qPWOqkDvcn7TJcBU4BIRmRrZqobMH4HF3R67EXjJGDMJeMm5H2v8wDeNMVOAcuBa53cc6+veApxljDkVKAUWO1NRx/p6t/sasLnL/XhZ70XGmNIuY88Htd5RFeh0Ob+pMaYVaD+/acwxxqwCDnV7+ALgXuf2vcCFw1nTcDDG7DfGrHNu12P/yQuJ8XU3VoNz1+tcDDG+3gAiMhZYCvyuy8Mxv969GNR6R1ug93bu0ngxuv3EIc51XoTrGVIiUgzMAt4kDtbd6XZYD1QBLxhj4mK9gV8B/xcIdnksHtbbAM+LyFrnfMswyPUOaT70ESSkc5eq6CciqcBfgOuMMUdEevrVxxZjTAAoFZEM4AkRmR7hkoaciHwCqDLGrBWRhREuZ7jNM8bsE5E84AUR2TLYN4y2Fnq8n7u0UkQKAJzrqgjXMyRExIsN8weMMY87D8fFugMYY2qBldh9KLG+3vOAT4nILmwX6lkicj+xv94YY/Y511XAE9gu5UGtd7QFeijnN41lTwJXOLevAP4WwVqGhNim+O+BzcaYX3R5KqbXXURynZY5IpIEnANsIcbX2xjzX8aYscaYYuz/88vGmMuI8fUWkRQRSWu/DZwLbGSQ6x11R4qKyPnYPrf285v+JLIVDQ0ReQhYiJ1OsxL4AfBX4FGgCPgI+KwxpvuO06gmIh8HVgPv0dmn+m1sP3rMrruIzMTuBHNjG1qPGmN+LCLZxPB6d+V0uXzLGPOJWF9vEZmAbZWD7fp+0Bjzk8Gud9QFulJKqZ5FW5eLUkqpXmigK6VUjNBAV0qpGKGBrpRSMUIDXSmlYoQGulJKxQgNdKWUihH/H/ZdR3NB6OITAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAv9UlEQVR4nO3deXiU1fXA8e/JAiEhQDYWE0IiBpFFFgOigGAVBbWiIgqidavUWq3+umntYq3aWpcu1oWiorhBLUpFiyuyKaIECLtA2JKwhiwESEK2+/vjTnBIJskkmSUzOZ/nyZPMvHfe97yBnLlzVzHGoJRSKvCF+DsApZRSnqEJXSmlgoQmdKWUChKa0JVSKkhoQldKqSAR5q8Lx8fHm5SUFH9dXimlAtLq1asPG2MSXB3zW0JPSUkhIyPDX5dXSqmAJCJ76jumTS5KKRUkNKErpVSQ0ISulFJBQhO6UkoFCU3oSikVJBpN6CIyS0QOicjGeo6LiDwjIlkisl5Ehno+TKWUUo1xp4b+KjC+geMTgDTH13TghZaHpZRSqqkaHYdujFkmIikNFJkIvGbsOrwrRaSLiPQwxuz3VJDKe6qqDfuKSskuKGFPfgnFZRWk94phUM8uhIf6p0XOGEP+8XL25JeQXXCc3IJSKqqqPXLudmEh9IyNJDk2kl5xUcREhiMip1y7sKSCPfnHyS4oIaeghPJKz1xbqRrpKbFc0Mfl3KAW8cTEokQgx+lxruO5OgldRKZja/EkJyd74NKqqYwxLFi3j3fX7CW7oITcwhIqququiR/dPozzescxuk8CF6TF0ysuCmMMBcfL2VNQQna+fQPIKSyhrKLKI7GVV1aTU1hKdv5xjpefek6nnNsitZf/j24fRnJcJN07RXCguIzs/BKOnqj0yrWVqnHnmN6tNqG7+u/uctcMY8xMYCZAenq67qzhpsLj5cxbncuibw9yZrdoRqclMKJ3HB3bN+2f7/CxE/xm/gY+3nSQ1PgozuoRzaX9u9MrLpJesZEkx0US2S6Mr3fms2z7YZZty+OTzQcB6N4pgmMnKjlWK9l169SeqCbGUZ+wECEpJpJzU2NtTHGRJMdGkRTTgYjwUI9co6yi6uSnkZpa+J78EvYWldK9cwTpvWJIjouiV6y9fs/YSI9dWylv88RfYi7Q0+lxErDPA+dt04wxrM0p4o2Ve/hg/X7KK6s5s1s0b2fkMvurPYSFCEN7xXBBWjyj0xIYkNiZ0JD6q5IfbTzAb+Zv4GhZJb+e0Jcfjj693vITBvZgwsAeGGPYnV/C8u15ZOwuJCYyPOCTXUR4KH26RdOnW7S/Q1HK4zyR0BcAd4vIXOBc4Ii2nzfficoq3lm9lzdW7mHz/mI6tg/j+vSeTBuRTN/unThRWcXqPYUs336Y5dvzeOqTbTz1yTa6RIYz8ox4LkiLZ1RaAoldOgBwpLSChxds4t21exmQ2Ik51w12O5mJCKnxUaTGR/GD81K8eNdKKU+QxvYUFZE5wFggHjgIPASEAxhjZojtUXoWOxKmBLjVGNPoqlvp6elGF+c61cHiMqa/vpp1OUWc1aMTN45IZuLgxAabVvKPneCLrMN8sf0wy7bncbD4BACnJ0Rxfu84Fm05xKGjJ/jJhWdwz/fO8FtHp1LKM0RktTEm3eUxf20SrQn9VOtyipj+egZHyyp5avIgJgzofsroC3cYY8g6dIxljtr7yp35JMVE8vTkQQzq2cU7gSulfKqhhO635XPVd+avzeX+dzbQNbo97951Pn27d2rWeUSEtG7RpHWL5vZRqVRWVRMaIk1+Y1BKBSZN6H5UVW144uNv+dfSnZybGssLN55DbFQ7j50/TJtXlGpTNKH7gTGG/UfK+M38DSzemseNI5J56Pv9tX1bKdUimtC9qLyymjXZhWw7eNQx7tnOfMwuKKGsopqwEOHRqwZw44he/g5VKRUENKF7kDGGHXnHWb49j+XbD7NyZz4ljhmPEeEh9IqNIjk2itFpCfSKi2R4amyz28uVUqo2TegeUFVt+MtH3/K/9fvZW1QKQEpcJJOGJjE6LZ7BPbuQEN1eOyeVUl6lCd0DFqzby8xlO/le367cdWFvRp+RQHJcpL/DUkq1MZrQW6i8spq/fbqd/qd14qUfpBPSwPR7pZTyJh1W0UJvZ+SQXVDCLy49U5O5UsqvNKG3QFlFFf/8fDvDUmIY64WlMJVSqik0obfA61/t4WDxCX5xyZna4amU8jtN6M10tKyC55dkcUGfBM49Pc7f4SillCb05nr5i10UllTwy0vO9HcoSikFaEJvloLj5by0fBcTBnRnYFJnf4ejlFKAJvRmmbF0ByXllfxsXB9/h6KUUidpQm+ig8VlzF6xm6uHJJGm25gppVoRTehN9M/Pt1NtDPddnObvUJRS6hRuJXQRGS8iW0UkS0QecHE8RkTmi8h6EflGRAZ4PlT/yy0sYe43OUwdnkzPWJ3ar5RqXRpN6CISCjwHTAD6AVNFpF+tYg8CmcaYs4EfAP/wdKCtwdJteVRWG24dmervUJRSqg53aujDgSxjzE5jTDkwF5hYq0w/YBGAMeZbIEVEunk00lYgM7uI2Kh2pOjCW0qpVsidhJ4I5Dg9znU852wdcA2AiAwHegFJtU8kItNFJENEMvLy8poXsR+tzSlicM8uOitUKdUquZPQXWUvU+vx40CMiGQC9wBrgco6LzJmpjEm3RiTnpAQWGufFJdVsCPvGIN7dvF3KEop5ZI7y+fmAj2dHicB+5wLGGOKgVsBxFZfdzm+gsb6nCMYgyZ0pVSr5U4NfRWQJiKpItIOmAIscC4gIl0cxwB+CCxzJPmgkZlTCMAgTehKqVaq0Rq6MaZSRO4GPgZCgVnGmE0icqfj+AzgLOA1EakCNgO3ezFmv8jMKeL0hCg6dwj3dyhKKeWSWzsWGWMWAgtrPTfD6eevgKCdaWOMITOniDF9uvo7FKWUqpfOFHVDbmEph4+VMzi5i79DUUqpemlCd0NmThEAQ7T9XCnVimlCd8Pa7CLah4VwZnddjEsp1XppQndDZk4hAxM7Ex6qvy6lVOulGaoR5ZXVbNxXrOPPlVKtnib0Rnx7oJjyymrtEFVKtXqa0BtR0yGqNXSlVGunCb0RmdlFJES3J7FLB3+HopRSDdKE3ohMXWFRKRUgNKE3oKiknJ2Hj2tzi1IqIGhCb4BOKFJKBRJN6A3IzClCBAYmdfZ3KEop1ShN6A3IzCkirWtHoiN0hUWlVOunCb0exhjWOTpElVIqEGhCr8ee/BIKSyoYkhzj71CUUsotmtDroROKlFKBxq2ELiLjRWSriGSJyAMujncWkfdFZJ2IbBKRWz0fqm9l5hQR2S6UPt10hUWlVGBoNKGLSCjwHDAB6AdMFZF+tYr9BNhsjBkEjAWedtpjNCCtzSliYGJnQkN0QpFSKjC4U0MfDmQZY3YaY8qBucDEWmUMEC12OmVHoACo9GikPlRWUcXmfUd0QS6lVEBxJ6EnAjlOj3Mdzzl7FrtR9D5gA3CvMaa69olEZLqIZIhIRl5eXjND9r7N+4upqDI6oUgpFVDcSeiu2hxMrceXApnAacBg4FkR6VTnRcbMNMakG2PSExISmhiq72RmFwEwuKeOcFFKBQ53Enou0NPpcRK2Ju7sVuBdY2UBu4C+ngnR99bmFNGjcwTdO0f4OxSllHKbOwl9FZAmIqmOjs4pwIJaZbKBiwBEpBtwJrDTk4H60po9hQzV8edKqQDTaEI3xlQCdwMfA1uAt40xm0TkThG501HsEeB8EdkALALuN8Yc9lbQ3nSouIy9RaUM0Q5RpVSACXOnkDFmIbCw1nMznH7eB1zi2dD8Y012IQBDe2kNXSkVWHSmaC1rsotoFxpC/9Pq9OkqpVSrpgm9lrXZhfRP7ET7sFB/h6KUUk2iCd1JeWU163OPaIeoUiogaUJ3smV/MScqqzWhK6UCkiZ0J991iHbxbyBKKdUMmtCdrMkuonunCHp07uDvUJRSqsk0oTtZm12otXOlVMDShO5w6GgZuYWl2n6ulApYmtAd1uwpAtAt55RSAUsTusPa7ELCQ0UnFCmlApYmdIc12YX0P60zEeE6oUgpFZg0oQMVVTqhSCkV+DSh4zShSEe4KKUCmCZ07PrngNbQlVIBTRM6dkJRt07t6aE7FCmlApgmdGyH6NDkGERcbZ+qlFKBwa2ELiLjRWSriGSJyAMujv9SRDIdXxtFpEpEYj0frufphCKlVLBoNKGLSCjwHDAB6AdMFZF+zmWMMU8aYwYbYwYDvwaWGmMKvBCvx63NLgJ0QS6lVOBzp4Y+HMgyxuw0xpQDc4GJDZSfCszxRHC+sObkhKLO/g5FqeBjDJQf93cU7is/DpUn/B1Fs7mT0BOBHKfHuY7n6hCRSGA88E7LQ/ONtXuK6KcTipTyrLJi+OZFeOF8+NNpsOgRqK72d1QNy/4a/jHYxnx4u7+jaRZ3ErqrnkJTT9nvA1/W19wiItNFJENEMvLy8tyN0WsqqqpZv7eIocld/B2KUsFh/3p4/154ui8s/AWEtoOzroTlT8G/b4QTR/0doWtr34DZV0C7KCgtghcvgu2f+juqJgtzo0wu0NPpcRKwr56yU2igucUYMxOYCZCenl7fm4LPfLv/KGUVukORChI7l8D+dTDiJxDqzp+2h1SUwab5kPEy5K6CsAgYcC0Muw0Sz7HNLt/MhI9+DS+Ng6lzIDbV9bmKcmD9XOgQAwOvg4hG1laqLIctC+DQZuh/DXQf0LTYqyrh09/Byufh9LFw7Su22WXuVHjrOhj3RzjvbnA1Aq6qErYuhB2fg6mqe7xdNIz6P+iY0LSYWsCdf/VVQJqIpAJ7sUn7htqFRKQzMAa40aMRetF3OxRpQlcBzBj46ln49PdgqiFrEUx+FSK9PNAsfwdkzILMN6G0EOLS4NI/w+CpNiHXEIFzfwQJZ8LbN8OLF8Lk2XD6GHu8usrGnPEybP/E3gPAJ7+Hs6+D9Nugx9mnXrtwD6x+Fda+Dscdn/aXPw09z4X026HfRAhvZF5JSQHMuw12LoZzfwyXPGrfCCNj4baP4b8/hk9+Cwc2wvf/8d35ivfB6tmwZjYc3Q/tO0O7yLrnP3YIivfCdbOb/KttLjGm8YqyiFwG/B0IBWYZYx4TkTsBjDEzHGVuAcYbY6a4c+H09HSTkZHRzLA946dz1rJyZz5fP3iRjkH3pepqCPHAFIjqKpAQ17Wn1siYpsVa87fZ0GsqyuCD+2DdHNu00ft78OGvoFOirQl3Pat5sVaWQ0WJq6Bg13KbyHcuhpAw6Hu5TaKpFzR+fwU7Yc5U20Y97o9QVQ6rX4GibIjqCkN/AOfcbJP0qlmwcR5UlkHSMHuNDjH22ts/sdfqM94+f9oQ+zvImAUFO6BDLAyZBkN+AB271o2jKBv+c7P9RHDF32DoTS5u1cCyJ2HxY/aTxqj/g3VzYeuH9k3njIvsm03apa4/ES19EhY/Cje+A2dc7Nav3R0istoYk+7ymDsJ3Rv8ndCrqw3D//QZ5/eO55mpQ/wWR5uz/GlYOQOufdkmgObau9rW9nqeC5Near1JvaLUNkesehkOrLdJd9jtkHxe/THvW2sT04Z5EJNqmy5cNT8cPQBzp8HeDBj7IFzwS/tGmbMK/j0Nyktg0otw5gT34z24yca6/m0ob6C9u1MSnHOLTYTR3d0/P9gO03fvgG0f2ccpo21i7HsFhLU7tWxpIWTOsbX3/Cz7XMduNvEPvRm69Dy1fHU17Fpqy3+70HVTSI2oBLj+DUge0XC8W96Hd38EFcchMg6G3Ajn3Fp/s1GNyhPw/HmAgR9/1fgnBjdpQndh494jXPHPL/jrdYO4ZmiS3+JoU0oL4W8D7R8GAhP+AsN+2PRkvP5teO9u21Z74ghc+FsY80uvhNxsh7NszXPtG1BWBPF9oOdw2Py+jTnhLJvEBl0PEZ1t8t30rk2m+9ZAeKRN/oc22zeCdh1h4GT7ZtB9oH1DmzvNJserZ0C/K0+9/pG9MPcG26Z+0e9t7bK+33PlCdj8nr12zkoIbQ8DroHuZ7suH3u6rXG2pJ2+ugq2fQxxvW1TTGOMgd1f2E7VtHEQGt74a4r32TbuyvK6x0JC4azvQ6fT3Is3fwfkbbW18rD27r0GYMdieP0qGPtrGFtnTmazaEJ34bnFWTz58VZW/eZiEqKb8A/UGq1/29YyuiS37DxH9kLWpzDkJvsf3tOWPA5L/gy3fgRf/t3W0M65BSY8Wbdm5kp1FSx6GL78B/QaCde9Bh8/COv/DVPesh/9/S17JSz+k60lhoTZpJF+O6SMsgm1/DhsfNfWIPethfAo6H0h7F4OZUcgoa9N9GdfDx262ES2d40tv/Ed2/xw2lBbk47uBlPm1N8RWFFq3/g2zoM+E6Bb/7plyo7YN5KSfJuo02+DwdO83/7elsy7DbZ8AHd9Zd/AWkgTugvXzfiKkopKPrhntN9i8IgDG2HGSOh/te0Ia67sr+3H9ON5cNlTMPwOj4UI2Jrk3wdAr1Ew9S2bnD9/FL74KySfD9e/DlHxDbz+CLzzQ9t2mn4bjP+LfROoKIVXLoPD2+CHnzW/zdgTVr1s2687doP0W237bXS3+svXJOptn0DqaJv4e51ff026pMC24a59AzonwVXPN/w7A/uG8OXfYekTrifMhIRBn0scbeBjPNO3oU5VvB+eHWY/od34ToubBzWh13KktIKhj3zKnWNO55eX9vVLDB7zv5/DqpfsH+bPtrjuAGrMmtfhg/+z7ZFRCXDoW7h7VcPJqKmWPw2L/gjTl9gOrBob5sF7P7EdYlPfss0JteXvgDlTbIdaTTONs+J9MHOsbaa443Pf1y6rKuDD+21yTrvEtulH6Mxj5WTlC/DRA/ZTZb+GJto3rqGE7sPBqq3HiqzDVFUbxvRpRvJrTU4cg3X/th1s2V/ZIVyjf+7+66sq7bCsr1+A0y+Eya/A8Xx44Tz7/KQXG3798Xx4c5Lt5b/w1/WXKz8OXz0HZ4w7NZkDDLzWftSfOw1mjHbdNlpVYUc33DTfdUdqp9Ns59arl8O8W2HaO+617xoDuRm2A3LL+1DlogYrIfaa6bfbttvaTVHH8+1oid3LYeS9cNFD3mmuUoFt2B2w9k348AE7Eql9tFcu0yYT+tJteUS3D2NIoM8Q3TjPjkS4+GE7PGr1qzDyPvcSSkmBTX47l8CIu2DcIzYJdoix51j2hB3BUN9IlKoKm8j2rbVfsafbDj5XMmbZNtoxv3J9PHGorbmvfsW2EdcWEm6HoMWk1H8/PYfD5X+FBXfb8djj/1R/2RPHYMN/bI36wAbb4dj/KvvppLbyEtthOOd66NzTDqmraUo5uMl+cjh6EK6eWf/9KxUaBlf8FV4eZ/uSLn3MK5dpcwndGMPSbXmMSosnPDSA2wuNsW223QbYZJZ+u02wOz63NcmGHM6CtybbMbhXPlt3DO7on9mOxv/9HO780nWH5ce/sbXSK/9p23UX3APxaTY5O6sohS+fse2zPYfXH1N0t5aPAhh6k03QK5+zbcGde9Ytc3ib/VRTftT+7i7/q5280lCN6dLH7GiJVS/bdv8lj9tPJTuX2KGEt31oxykr1ZCew+1Qy5UvwOAbXHdSt1CbS+jbDx1j/5Ey7r3Id9NxvWLvGjuc7fKnbSdL38ttZ9yqlxtO6FWVMO8W28l4yweux+CGd7Ado29NtjMQR//s1ONrXoNv/mWnmA/9AZx5Gcy80DabTF9yatv7mtfg+CEY84on7rpxlz5m29pX/NP18dB2tgM5/Xb7B+ZOB1VouG337DfRTojJeAXWOdr7J78KnXp49BZUELv4D7Z5b81rtj/Iw9pcQl+61U4TvqBPgCf0jFl2yNvA6+zj0HA73PCLv9qad+0JFzVWvWRrsZNnNzyhos8ldqLH0idgwCSI6WWfz/4aPviZbXMf90f7XFQ8THkTZl0Kb98EN79vx+pWnoAv/m5HsaSM8titNyg0HKb9x455dyW8g/1qrvg025xz6WOtdzKTar0iY+1orNjTvXL6AG5zaJ6l2/Lo060jp3VpwR+1v5UW2jHJZ08+dfbgOTfbppg1r7l+XfF+22TQ+yL3etrHP26T1keOppAje+2KeZ2T4NpZp3Y89jjbDqPL+do21Rhj1/g4us/3k35E7B+Oq6+WJPPa11CqOeJ6e+3/T5tK6CXllXyzq4AxgV47X/dvqCy147GddUm2w+bWvGY7LWv75Dd27YzLnnTvP1SXnrZde+tCOxnm39Ps+h5T57geGtj/ajv9fO3rdvW6L/4Giem2Nq+U8ro2ldBX7synvKo6sIcrGmObWxLTocegusfTb4NjB+wCQs52LLa1+tE/a9pstRF32dmL826zo1mumdnw5J2xD9o29Y8ftAsgjfmV1maV8pE2ldCXbs2jQ3go6SkBvFzuni/h8Na6tfMaaePswkkZs757rvKE3WwgJtUOSWyK0HA7EiQkDL7328an14eEwNX/gq79IWm4/cSglPKJNtUpunRbHuf1jgvs7eYyZtlZiP2vdn08JNSuj7L4UTvDMq43rHjGrlQ37Z3mrfiWMhJ+tbPxzQZqRHSCHy2F6kqtnSvlQ22mhr778HF255cEdvv5sTzYvAAG3eB6Qf0aQ2+yNerVr0DBLlj2lO0ETWvBmszuJvMaoeGe64BUSrmlzdTQl223wxUDOqFnvgHVFXbhp4ZEd7dNI2vfhIObbXK/9M++iVEp5TdtJqEv3ZpHSlwkKfFR/g7FOn4Y8r5t2msyXrGbAbizfnT6bXbK+o5FcMlj0DmxeXEqpQKGWwldRMYD/8BuQfeSMeZxF2XGYrepCwcOG2PGeCzKFjpRWcWKHflcl+7njSyMgT0r7BoimxfY2nZTjXvYvXIpF9hNFULC7X6OSqmg12hCF5FQ4DlgHJALrBKRBcaYzU5lugDPY/cUzRaRVjUuMGN3IaUVVYw500/NLWVH7NjxjFmQt8VuKjvsh3Y2ZkgTPiSFRdi9Fd0REgK3/M+e353dXZRSAc+dbDIcyDLG7AQQkbnARGCzU5kbgHeNMdkAxphDng60JZZuy6NdaAgjTo/z7YWNsTMzVz5vJ+ScNsQuhjVgUsOdmp7SnLXRlVIBy52EngjkOD3OBc6tVaYPEC4iS4Bo4B/GmDrzz0VkOjAdIDm5hdulNcGybXkMT40lsp2PuwxWvgDLn4J+V9m1smuvRKiUUh7kToZzNZC49jZHYcA5wEVAB+ArEVlpjNl2youMmQnMBLtjUdPDbbrKqmq2HzrGRWf5uLa6Y7Gdat/3Crj2Fd3aSynlde4k9FzAeem+JGCfizKHjTHHgeMisgwYBGzDz/YfKaOq2tAzxgdNHDXyd8B/brFT5q+eoclcKeUT7mSaVUCaiKSKSDtgCrCgVpn3gNEiEiYikdgmmS2eDbV5cgpLAOgZ66OEfuIozL3BzpCc8pbXtppSSqnaGq2hG2MqReRu4GPssMVZxphNInKn4/gMY8wWEfkIWA9UY4c2bvRm4O7KLSgF8E0Nvboa3v2R3QThpnchNtX711RKKQe3egmNMQuBhbWem1Hr8ZPAk54LzTNyC0sIEejRpRlrmDTV0sdh6/9g/F/g9LHev55SSjkJ+pmiOYWl9OjcwXP7h5Yft2uK15a1CJb+BQbfqBN5lFJ+EfwJvaCEpBgPLRK1axnMvpK6g3wckobZnb11hUGllB8Ef0IvLGF0mgdmiBoDi/8E0T1g5E/rHg8Nh/7X2L00lVLKD4I6oZdVVHGw+IRnOkR3fwHZX8GEJ7RJRSnVKgX1AOm9RY4RLrEeaHJZ9gREdYWhP2j5uZRSyguCOqHnFHhoDHr217b9fORPddMGpVSrFdwJvdBDY9CXPQGRcfXv46mUUq1AUCf03IIS2oWG0DW6BR2Ve1dD1mdw3k+gXSvZHEMppVwI7oReWEpiTAdCQlowjHDZUxDRBYbd4bG4lFLKG4I6oecUtnAM+oENsHUhjPhx0zdJVkopHwvuhF5Q0rIO0WVPQbtoHaaolAoIQZvQj52opLCkovkdonlb7SbL506HDjGeDU4ppbwgaBP6d0MWm9nksuwpCI+EET/xYFRKKeU9wZ/Qm1NDz98BG+fBsNsgysf7kCqlVDMFb0KvGYPenDb0r56DkHA47x4PR6WUUt4TvAm9oITIdqHERIY37YUnjsL6f8OAayC6m3eCU0opL3AroYvIeBHZKiJZIvKAi+NjReSIiGQ6vn7v+VCbJrewlJ4xkUhTl7LdMA/Kj+msUKVUwGl0tUURCQWeA8ZhN4NeJSILjDGbaxVdboy5wgsxNktuYUnTO0SNgYxZ0G2AXdtcKaUCiDs19OFAljFmpzGmHJgLTPRuWC1jjHFsbNHE9vO9a+DAeki/VTepUEoFHHcSeiKQ4/Q41/FcbeeJyDoR+VBE+nskumYqLKngeHlV0ztEM2ZBu45w9vXeCUwppbzInQ0uXFVVa+/BtgboZYw5JiKXAf8F0uqcSGQ6MB0gOTm5aZE2wXdDFpvQ5FJaCBvfgUFToH20lyJTSinvcaeGngv0dHqcBOxzLmCMKTbGHHP8vBAIF5H42icyxsw0xqQbY9ITEjywLVw9cgqbsQ76urlQWWqbW5RSKgC5k9BXAWkikioi7YApwALnAiLSXRzDSURkuOO8+Z4O1l05BXYMutsLc9V0hiamQ49BXoxMKaW8p9EmF2NMpYjcDXwMhAKzjDGbROROx/EZwLXAj0WkEigFphhjajfL+ExOYQldIsOJjnBzDPqeL+HwNpj4vHcDU0opL3Jrk2hHM8rCWs/NcPr5WeBZz4bWfDVj0N2WMQsiOkP/q70XlFJKeVlQzhTNLWjCGPRjebB5AQy6Adq1cKs6pZTyo6BL6NXVpmk19Mw3oLpCO0OVUgEv6BL6oaMnKK+qJsmdES7V1ZDxCvQaBQlnej84pZTyoqBL6CeHLLozwmXn51C0xy6Tq5RSAS74EnpBE8agfz0TIuOh7/e9HJVSSnlfECZ0OwY9sUsjNfT962D7x3a/0LB2PohMKaW8K/gSemEJXaPbExEe2nDBZU9B+04wfLpvAlNKKS8LvoReUNJ4c8uhLbBlga2dd+jik7iUUsrbgi6h2yGLjTS3LHvKrqo44i7fBKWUUj4QVAm9oqqa/UdKG66hH86CTe/CsNshMtZ3wSmllJcFVULfX1RGtaHhSUXLn4bQ9roBtFIq6ARVQq8Zg55U37T/gl12A+j0W6Gj95bvVUopfwiuhH5yY4t6auhf/A1CwuD8n/owKqWU8o3gSuiFJYSGCD06R9Q9WJQDmW/B0JugUw/fB6eUUl4WXAm9oJQenSMIC3VxW1/+w34feZ9PY1JKKV8JroReWOK6ueXoAVjzGgyeCl161j2ulFJBIKgSem5hqet10L98BqorYdTPfB+UUkr5iFsJXUTGi8hWEckSkQcaKDdMRKpE5FrPheiesooq8o6eqFtDr6qEta/DgGsgNtXXYSmllM80mtBFJBR4DpgA9AOmiki/esr9Bbv3qM/lFtazyuLBDXCiGPqM90NUSinlO+7U0IcDWcaYncaYcmAuMNFFuXuAd4BDHozPbdmOIYvJcbUS+u4v7fde5/s4IqWU8i13EnoikOP0ONfx3EkikghcDcygASIyXUQyRCQjLy+vqbE2KDvfkdBr19D3rICYVOh0mkevp5RSrY07CV1cPGdqPf47cL8xpqqhExljZhpj0o0x6QkJnp2puaeghMh2ocRFOa1tXl0N2SsgZaRHr6WUUq1RmBtlcgHnsX5JwL5aZdKBuSICEA9cJiKVxpj/eiJId+QUlJAcG4kjBitvC5QWQi9N6Eqp4OdOQl8FpIlIKrAXmALc4FzAGHNy+IiIvAp84MtkDrYNvVdc1KlP7llhv2tCV0q1AY02uRhjKoG7saNXtgBvG2M2icidInKntwN0hzGGbEcN/RS7v4BOSdAl2T+BKaWUD7lTQ8cYsxBYWOs5lx2gxphbWh5W0+QdO0FZRfWpCd0YW0PvfSGIq24ApZQKLkExU7RmlcVTEnp+Fhw/pMMVlVJtRlAk9Jox6KdMKtpTM/58lB8iUkop3wuOhJ5fCkCS816iu7+EqK4Q19tPUSmllG8FR0IvKKF7pwgiwkPtE8bYGnrKSG0/V0q1GUGR0HMKSk6d8l+0B4r36nBFpVSbEhQJvc6QRR1/rpRqgwI+oZdVVHGguOzUhL77S+gQAwl9/ReYUkr5WMAn9Jplc5Nrj3DpNRJCAv72lFLKbQGf8eoMWSzeB4W7dPy5UqrNCfyEXnvZXG0/V0q1UYGf0AtK6RAeSnxHx7K5u7+AdtHQfaB/A1NKKR9zay2X1iy79rK5e1ZA8ggICfVvYEopKioqyM3NpayszN+hBJyIiAiSkpIIDw93+zUBn9BzCkq+az8/lgeHt8Lgqf4NSikFQG5uLtHR0aSkpJy6V4FqkDGG/Px8cnNzSU11f3P7gG5yqbNsbra2nyvVmpSVlREXF6fJvIlEhLi4uCZ/sgnohH74WDmlFVUkxzrWcNn9JYRHQo/Bfo1LKfUdTebN05zfW0An9Johiyd3KtqzApKGQVi7Bl6llFLBya2ELiLjRWSriGSJyAMujk8UkfUikikiGSLikzVrc5zHoJcUwMGNkKLL5SqlrKKiIp5//vlmvfayyy6jqKjIswF5WaMJXURCgeeACUA/YKqI9KtVbBEwyBgzGLgNeMnDcbpUU0NPiukAW94HDKSN88WllVIBoKGEXlVV1eBrFy5cSJcuXbwQlfe4M8plOJBljNkJICJzgYnA5poCxphjTuWjAOPJIOuzJ99p2dyN8yDuDG0/V6qVevj9TWzeV+zRc/Y7rRMPfb9/vccfeOABduzYweDBgxk3bhyXX345Dz/8MD169CAzM5PNmzdz1VVXkZOTQ1lZGffeey/Tp08HICUlhYyMDI4dO8aECRMYNWoUK1asIDExkffee48OHTqccq3333+fRx99lPLycuLi4njzzTfp1q0bx44d45577iEjIwMR4aGHHmLSpEl89NFHPPjgg1RVVREfH8+iRYta/PtwJ6EnAjlOj3OBc2sXEpGrgT8DXYHLXZ1IRKYD0wGSk1u+cXNOzQiX4v2wazmMuV/XP1dKnfT444+zceNGMjMzAViyZAnffPMNGzduPDkccNasWcTGxlJaWsqwYcOYNGkScXFxp5xn+/btzJkzhxdffJHrrruOd955hxtvvPGUMqNGjWLlypWICC+99BJPPPEETz/9NI888gidO3dmw4YNABQWFpKXl8cdd9zBsmXLSE1NpaCgwCP3605Cd5Uh69TAjTHzgfkicgHwCHCxizIzgZkA6enpLa7FZxeUMPKMeNg034Y08NqWnlIp5SUN1aR9afjw4aeM7X7mmWeYP38+ADk5OWzfvr1OQk9NTWXw4MEAnHPOOezevbvOeXNzc7n++uvZv38/5eXlJ6/x2WefMXfu3JPlYmJieP/997ngggtOlomNjfXIvbnTKZoL9HR6nATsq6+wMWYZ0FtE4lsYW4NOWTZ34zzofjbEp3nzkkqpIBAVFXXy5yVLlvDZZ5/x1VdfsW7dOoYMGeJy7Hf79u1P/hwaGkplZWWdMvfccw933303GzZs4F//+tfJ8xhj6gxBdPWcJ7iT0FcBaSKSKiLtgCnAAucCInKGOKITkaFAOyDf08E6yy20+4ieFZEHe1fDwMnevJxSKgBFR0dz9OjReo8fOXKEmJgYIiMj+fbbb1m5cmWzr3XkyBESExMBmD179snnL7nkEp599tmTjwsLCznvvPNYunQpu3btAvBYk0ujCd0YUwncDXwMbAHeNsZsEpE7ReROR7FJwEYRycSOiLneGOPVjtGaIYv98z+1Twy4xpuXU0oFoLi4OEaOHMmAAQP45S9/Wef4+PHjqays5Oyzz+Z3v/sdI0aMaPa1/vCHPzB58mRGjx5NfPx3DRS//e1vKSwsZMCAAQwaNIjFixeTkJDAzJkzueaaaxg0aBDXX399s6/rTLycd+uVnp5uMjIymv362St289CCjWQl/pGwqDi47UMPRqeU8oQtW7Zw1lln+TuMgOXq9yciq40x6a7KB+xM0eyCEgaF7yUsfysMnOTvcJRSyu8COqFPjfwGJBT6XeXvcJRSyu8CNqHn5B/n4qrl0Pt7EOXVATVKKRUQAnI9dGMMsYWZxIcc1LHnSinlEJA19MPHyrmk+gsqQ9pDX5eTUpVSqs0JyISefbiYy0NXUpB4IbSP9nc4SinVKgRkQi/btoQEKaaqv45uUUp5VseOHf0dQrMFZEKP2fkexaYDMYO0uUUppWoEXqdoRRmpeZ+zKHQEV3SIary8Uqp1+PABOLDBs+fsPhAmPF7v4fvvv59evXpx1113AXY2Z3R0ND/60Y+YOHEihYWFVFRU8OijjzJx4sQGL1XfMruulsGtb8lcbwu8hJ71KR2qj7M25iKu8HcsSqlWbcqUKdx3330nE/rbb7/NRx99REREBPPnz6dTp04cPnyYESNGcOWVVza4YJarZXarq6tdLoPraslcXwi8hN5jEC+ETuNI9/P8HYlSqikaqEl7y5AhQzh06BD79u0jLy+PmJgYkpOTqaio4MEHH2TZsmWEhISwd+9eDh48SPfu3es9l6tldvPy8lwug+tqyVxfCLiEXhaVyBMll3NvXCd/h6KUCgDXXnst8+bN48CBA0yZMgWAN998k7y8PFavXk14eDgpKSkul82t4bzMbmRkJGPHjqWsrKzeZXC9tTxuYwKuU3RvUSnGYNdBV0qpRkyZMoW5c+cyb948rr3WTkQ8cuQIXbt2JTw8nMWLF7Nnz54Gz1HfMrv1LYPraslcXwi4hF6zMXSvOE3oSqnG9e/fn6NHj5KYmEiPHj0AmDZtGhkZGaSnp/Pmm2/St2/fBs9R3zK79S2D62rJXF8IuOVzM3YXMHPZTv50zUDiO7Zv/AVKKb/R5XNbpqnL5wZcG3p6SizpKZ7Zf08ppYJJwDW5KKWUcs2thC4i40Vkq4hkicgDLo5PE5H1jq8VIjLI86EqpQKRv5p1A11zfm+NJnQRCcXuEzoB6AdMFZF+tYrtAsYYY84GHgFmNjkSpVTQiYiIID8/X5N6ExljyM/PJyIiokmvc6cNfTiQZYzZCSAic4GJwGani69wKr8SSGpSFEqpoJSUlERubi55eXn+DiXgREREkJTUtFTqTkJPBHKcHucC5zZQ/nbA5Y7NIjIdmA6QnJzsZohKqUAVHh5+chal8j532tBdTXdy+flJRC7EJvT7XR03xsw0xqQbY9ITEhLcj1IppVSj3Kmh5wI9nR4nAftqFxKRs4GXgAnGmHzPhKeUUspd7tTQVwFpIpIqIu2AKcAC5wIikgy8C9xkjNnm+TCVUko1xq2ZoiJyGfB3IBSYZYx5TETuBDDGzBCRl4BJQM2CCJX1zWRyOmeeU/mmigcON/O1ga6t3rved9ui912/XsYYl23Wfpv63xIiktHYG0awaqv3rvfdtuh9N4/OFFVKqSChCV0ppYJEoCb0tjwTta3eu95326L33QwB2YaulFKqrkCtoSullKpFE7pSSgWJgEvojS3lGyxEZJaIHBKRjU7PxYrIpyKy3fHdN1uJ+5CI9BSRxSKyRUQ2ici9jueD+t5FJEJEvhGRdY77ftjxfFDfdw0RCRWRtSLygeNx0N+3iOwWkQ0ikikiGY7nWnTfAZXQ3VzKN1i8Coyv9dwDwCJjTBqwyPE42FQCPzfGnAWMAH7i+DcO9ns/AXzPGDMIGAyMF5ERBP9917gX2OL0uK3c94XGmMFOY89bdN8BldBxWsrXGFMO1CzlG3SMMcuAglpPTwRmO36eDVzly5h8wRiz3xizxvHzUewfeSJBfu/GOuZ4GO74MgT5fQOISBJwOXYtqBpBf9/1aNF9B1pCd7WUb6KfYvGHbsaY/WATH9DVz/F4lYikAEOAr2kD9+5odsgEDgGfGmPaxH1jlxX5FVDt9FxbuG8DfCIiqx1Li0ML7zvQNol2eylfFdhEpCPwDnCfMaZYxNU/fXAxxlQBg0WkCzBfRAb4OSSvE5ErgEPGmNUiMtbP4fjaSGPMPhHpCnwqIt+29ISBVkN3aynfIHZQRHoAOL4f8nM8XiEi4dhk/qYx5l3H023i3gGMMUXAEmwfSrDf90jgShHZjW1C/Z6IvEHw3zfGmH2O74eA+dgm5Rbdd6Al9EaX8g1yC4CbHT/fDLznx1i8QmxV/GVgizHmr06HgvreRSTBUTNHRDoAFwPfEuT3bYz5tTEmyRiTgv17/twYcyNBft8iEiUi0TU/A5cAG2nhfQfcTFFXS/n6NyLvEJE5wFjscpoHgYeA/wJvA8lANjDZGFO74zSgicgoYDmwge/aVB/EtqMH7b07NoiZjf1/HQK8bYz5o4jEEcT37czR5PILY8wVwX7fInI6tlYOtun7Lcey5C2674BL6EoppVwLtCYXpZRS9dCErpRSQUITulJKBQlN6EopFSQ0oSulVJDQhK6UUkFCE7pSSgWJ/wfklnKcJsvntAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the loss\n",
    "plt.plot(r.history['loss'], label='train loss')\n",
    "plt.plot(r.history['val_loss'], label='val loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('LossVal_loss')\n",
    "\n",
    "# plot the accuracy\n",
    "plt.plot(r.history['accuracy'], label='train acc')\n",
    "plt.plot(r.history['val_accuracy'], label='val acc')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('AccVal_acc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b847207a",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2876ba05",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "c4f1a24e",
   "metadata": {},
   "source": [
    "#### InceptionV3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "e8e3d43e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/inception_v3/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5\n",
      "87916544/87910968 [==============================] - 12s 0us/step\n",
      "87924736/87910968 [==============================] - 12s 0us/step\n",
      "Found 300 images belonging to 5 classes.\n",
      "Found 100 images belonging to 5 classes.\n",
      "Epoch 1/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 13.5134 - accuracy: 0.1967\n",
      "Epoch 00001: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 17s 987ms/step - loss: 13.5134 - accuracy: 0.1967 - val_loss: 7.2028 - val_accuracy: 0.4000\n",
      "Epoch 2/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 7.7936 - accuracy: 0.3967\n",
      "Epoch 00002: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 549ms/step - loss: 7.7936 - accuracy: 0.3967 - val_loss: 5.4597 - val_accuracy: 0.4900\n",
      "Epoch 3/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 3.2103 - accuracy: 0.6000\n",
      "Epoch 00003: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 554ms/step - loss: 3.2103 - accuracy: 0.6000 - val_loss: 3.7328 - val_accuracy: 0.4500\n",
      "Epoch 4/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.6978 - accuracy: 0.7033\n",
      "Epoch 00004: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 543ms/step - loss: 1.6978 - accuracy: 0.7033 - val_loss: 3.5716 - val_accuracy: 0.4300\n",
      "Epoch 5/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 1.2069 - accuracy: 0.7533\n",
      "Epoch 00005: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 546ms/step - loss: 1.2069 - accuracy: 0.7533 - val_loss: 2.5275 - val_accuracy: 0.5200\n",
      "Epoch 6/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.5504 - accuracy: 0.8600\n",
      "Epoch 00006: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 572ms/step - loss: 0.5504 - accuracy: 0.8600 - val_loss: 2.6891 - val_accuracy: 0.5200\n",
      "Epoch 7/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3860 - accuracy: 0.8933\n",
      "Epoch 00007: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 566ms/step - loss: 0.3860 - accuracy: 0.8933 - val_loss: 2.7838 - val_accuracy: 0.5100\n",
      "Epoch 8/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3596 - accuracy: 0.8967\n",
      "Epoch 00008: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 539ms/step - loss: 0.3596 - accuracy: 0.8967 - val_loss: 3.5368 - val_accuracy: 0.4800\n",
      "Epoch 9/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2252 - accuracy: 0.9233\n",
      "Epoch 00009: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 539ms/step - loss: 0.2252 - accuracy: 0.9233 - val_loss: 2.3086 - val_accuracy: 0.5400\n",
      "Epoch 10/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3226 - accuracy: 0.8833\n",
      "Epoch 00010: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 613ms/step - loss: 0.3226 - accuracy: 0.8833 - val_loss: 2.4631 - val_accuracy: 0.5300\n",
      "Epoch 11/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2296 - accuracy: 0.9167\n",
      "Epoch 00011: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 550ms/step - loss: 0.2296 - accuracy: 0.9167 - val_loss: 3.7676 - val_accuracy: 0.5200\n",
      "Epoch 12/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1793 - accuracy: 0.9567\n",
      "Epoch 00012: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 543ms/step - loss: 0.1793 - accuracy: 0.9567 - val_loss: 3.2012 - val_accuracy: 0.5500\n",
      "Epoch 13/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1602 - accuracy: 0.9500\n",
      "Epoch 00013: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 542ms/step - loss: 0.1602 - accuracy: 0.9500 - val_loss: 3.0986 - val_accuracy: 0.4600\n",
      "Epoch 14/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1350 - accuracy: 0.9533\n",
      "Epoch 00014: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 528ms/step - loss: 0.1350 - accuracy: 0.9533 - val_loss: 3.0622 - val_accuracy: 0.5000\n",
      "Epoch 15/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0767 - accuracy: 0.9700\n",
      "Epoch 00015: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 503ms/step - loss: 0.0767 - accuracy: 0.9700 - val_loss: 2.6323 - val_accuracy: 0.5500\n",
      "Epoch 16/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0704 - accuracy: 0.9767\n",
      "Epoch 00016: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 562ms/step - loss: 0.0704 - accuracy: 0.9767 - val_loss: 2.4409 - val_accuracy: 0.5700\n",
      "Epoch 17/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0786 - accuracy: 0.9700\n",
      "Epoch 00017: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 554ms/step - loss: 0.0786 - accuracy: 0.9700 - val_loss: 3.1497 - val_accuracy: 0.5600\n",
      "Epoch 18/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1500 - accuracy: 0.9533\n",
      "Epoch 00018: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 570ms/step - loss: 0.1500 - accuracy: 0.9533 - val_loss: 3.0563 - val_accuracy: 0.5300\n",
      "Epoch 19/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1051 - accuracy: 0.9433\n",
      "Epoch 00019: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 537ms/step - loss: 0.1051 - accuracy: 0.9433 - val_loss: 3.3127 - val_accuracy: 0.5400\n",
      "Epoch 20/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1043 - accuracy: 0.9700\n",
      "Epoch 00020: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 520ms/step - loss: 0.1043 - accuracy: 0.9700 - val_loss: 2.5545 - val_accuracy: 0.5500\n",
      "Epoch 21/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0578 - accuracy: 0.9833\n",
      "Epoch 00021: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 554ms/step - loss: 0.0578 - accuracy: 0.9833 - val_loss: 2.8898 - val_accuracy: 0.5700\n",
      "Epoch 22/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1441 - accuracy: 0.9467\n",
      "Epoch 00022: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 6s 552ms/step - loss: 0.1441 - accuracy: 0.9467 - val_loss: 3.6358 - val_accuracy: 0.5200\n",
      "Epoch 23/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1426 - accuracy: 0.9467\n",
      "Epoch 00023: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 546ms/step - loss: 0.1426 - accuracy: 0.9467 - val_loss: 3.3384 - val_accuracy: 0.5300\n",
      "Epoch 24/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0773 - accuracy: 0.9767\n",
      "Epoch 00024: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0773 - accuracy: 0.9767 - val_loss: 3.3790 - val_accuracy: 0.5600\n",
      "Epoch 25/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0884 - accuracy: 0.9733\n",
      "Epoch 00025: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 11s 1s/step - loss: 0.0884 - accuracy: 0.9733 - val_loss: 3.4407 - val_accuracy: 0.5400\n",
      "Epoch 26/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0914 - accuracy: 0.9667\n",
      "Epoch 00026: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 992ms/step - loss: 0.0914 - accuracy: 0.9667 - val_loss: 3.1864 - val_accuracy: 0.5100\n",
      "Epoch 27/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1346 - accuracy: 0.9533\n",
      "Epoch 00027: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 969ms/step - loss: 0.1346 - accuracy: 0.9533 - val_loss: 3.0331 - val_accuracy: 0.5900\n",
      "Epoch 28/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1099 - accuracy: 0.9700\n",
      "Epoch 00028: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.1099 - accuracy: 0.9700 - val_loss: 3.2851 - val_accuracy: 0.5400\n",
      "Epoch 29/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10/10 [==============================] - ETA: 0s - loss: 0.0302 - accuracy: 0.9933\n",
      "Epoch 00029: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0302 - accuracy: 0.9933 - val_loss: 3.2118 - val_accuracy: 0.5300\n",
      "Epoch 30/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0386 - accuracy: 0.9767\n",
      "Epoch 00030: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0386 - accuracy: 0.9767 - val_loss: 3.0937 - val_accuracy: 0.5700\n",
      "Epoch 31/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0212 - accuracy: 0.9900\n",
      "Epoch 00031: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0212 - accuracy: 0.9900 - val_loss: 2.5367 - val_accuracy: 0.5900\n",
      "Epoch 32/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0277 - accuracy: 0.9900\n",
      "Epoch 00032: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0277 - accuracy: 0.9900 - val_loss: 2.7025 - val_accuracy: 0.6200\n",
      "Epoch 33/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0091 - accuracy: 0.9933\n",
      "Epoch 00033: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 995ms/step - loss: 0.0091 - accuracy: 0.9933 - val_loss: 3.2072 - val_accuracy: 0.5700\n",
      "Epoch 34/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0215 - accuracy: 0.9867\n",
      "Epoch 00034: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0215 - accuracy: 0.9867 - val_loss: 2.8941 - val_accuracy: 0.5800\n",
      "Epoch 35/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0105 - accuracy: 0.9967\n",
      "Epoch 00035: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 990ms/step - loss: 0.0105 - accuracy: 0.9967 - val_loss: 3.1322 - val_accuracy: 0.6200\n",
      "Epoch 36/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0074 - accuracy: 0.9967\n",
      "Epoch 00036: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 1s/step - loss: 0.0074 - accuracy: 0.9967 - val_loss: 2.5332 - val_accuracy: 0.6900\n",
      "Epoch 37/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0294 - accuracy: 0.9933\n",
      "Epoch 00037: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 971ms/step - loss: 0.0294 - accuracy: 0.9933 - val_loss: 2.7480 - val_accuracy: 0.6000\n",
      "Epoch 38/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0245 - accuracy: 0.9900\n",
      "Epoch 00038: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 983ms/step - loss: 0.0245 - accuracy: 0.9900 - val_loss: 2.7312 - val_accuracy: 0.6000\n",
      "Epoch 39/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0258 - accuracy: 0.9933\n",
      "Epoch 00039: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 10s 986ms/step - loss: 0.0258 - accuracy: 0.9933 - val_loss: 2.5274 - val_accuracy: 0.6000\n",
      "Epoch 40/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0053 - accuracy: 1.0000\n",
      "Epoch 00040: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 497ms/step - loss: 0.0053 - accuracy: 1.0000 - val_loss: 2.5965 - val_accuracy: 0.5800\n",
      "Epoch 41/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0095 - accuracy: 0.9967\n",
      "Epoch 00041: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 502ms/step - loss: 0.0095 - accuracy: 0.9967 - val_loss: 2.5018 - val_accuracy: 0.6000\n",
      "Epoch 42/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0120 - accuracy: 0.9967\n",
      "Epoch 00042: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 476ms/step - loss: 0.0120 - accuracy: 0.9967 - val_loss: 2.5779 - val_accuracy: 0.5800\n",
      "Epoch 43/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0179 - accuracy: 0.9933\n",
      "Epoch 00043: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 469ms/step - loss: 0.0179 - accuracy: 0.9933 - val_loss: 2.6207 - val_accuracy: 0.6100\n",
      "Epoch 44/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0124 - accuracy: 0.9967\n",
      "Epoch 00044: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 472ms/step - loss: 0.0124 - accuracy: 0.9967 - val_loss: 2.8043 - val_accuracy: 0.5700\n",
      "Epoch 45/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0072 - accuracy: 0.9967\n",
      "Epoch 00045: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 479ms/step - loss: 0.0072 - accuracy: 0.9967 - val_loss: 3.0669 - val_accuracy: 0.5300\n",
      "Epoch 46/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0312 - accuracy: 0.9867\n",
      "Epoch 00046: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 478ms/step - loss: 0.0312 - accuracy: 0.9867 - val_loss: 3.2654 - val_accuracy: 0.5000\n",
      "Epoch 47/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0926 - accuracy: 0.9733\n",
      "Epoch 00047: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 474ms/step - loss: 0.0926 - accuracy: 0.9733 - val_loss: 2.7566 - val_accuracy: 0.5700\n",
      "Epoch 48/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0213 - accuracy: 0.9933\n",
      "Epoch 00048: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.0213 - accuracy: 0.9933 - val_loss: 3.0900 - val_accuracy: 0.5300\n",
      "Epoch 49/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0218 - accuracy: 0.9867\n",
      "Epoch 00049: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 5s 461ms/step - loss: 0.0218 - accuracy: 0.9867 - val_loss: 2.3045 - val_accuracy: 0.6200\n",
      "Epoch 50/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0156 - accuracy: 0.9933\n",
      "Epoch 00050: saving model to best_inception_model.h5\n",
      "10/10 [==============================] - 4s 451ms/step - loss: 0.0156 - accuracy: 0.9933 - val_loss: 2.3451 - val_accuracy: 0.6200\n"
     ]
    }
   ],
   "source": [
    "r, model, test_set = create_model(InceptionV3,'best_inception_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "d4784e74",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 0s - loss: 2.3451 - accuracy: 0.6200 - 465ms/epoch - 116ms/step\n",
      "Restored model, accuracy: 62.00%\n"
     ]
    }
   ],
   "source": [
    "# Inception Model\n",
    "model.save('model_inception.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "99103ea9",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 2s - loss: 2.3451 - accuracy: 0.6200 - 2s/epoch - 449ms/step\n",
      "Restored model, accuracy: 62.00%\n"
     ]
    }
   ],
   "source": [
    "# Inception model while checkpointing\n",
    "model=load_model('best_inception_model.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "044769e4",
   "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": [
    "# plot the loss\n",
    "plt.plot(r.history['loss'], label='train loss')\n",
    "plt.plot(r.history['val_loss'], label='val loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('LossVal_loss')\n",
    "\n",
    "# plot the accuracy\n",
    "plt.plot(r.history['accuracy'], label='train acc')\n",
    "plt.plot(r.history['val_accuracy'], label='val acc')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('AccVal_acc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "172c12db",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "b89d88d5",
   "metadata": {},
   "source": [
    "#### Xception"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "6f940e21",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 300 images belonging to 5 classes.\n",
      "Found 100 images belonging to 5 classes.\n",
      "Epoch 1/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 8.2005 - accuracy: 0.3600\n",
      "Epoch 00001: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 8s 613ms/step - loss: 8.2005 - accuracy: 0.3600 - val_loss: 4.1534 - val_accuracy: 0.4500\n",
      "Epoch 2/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 3.2490 - accuracy: 0.5300\n",
      "Epoch 00002: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 541ms/step - loss: 3.2490 - accuracy: 0.5300 - val_loss: 4.8052 - val_accuracy: 0.4700\n",
      "Epoch 3/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 2.2076 - accuracy: 0.6200\n",
      "Epoch 00003: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 490ms/step - loss: 2.2076 - accuracy: 0.6200 - val_loss: 3.7647 - val_accuracy: 0.4700\n",
      "Epoch 4/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.8369 - accuracy: 0.7800\n",
      "Epoch 00004: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 507ms/step - loss: 0.8369 - accuracy: 0.7800 - val_loss: 2.1314 - val_accuracy: 0.5900\n",
      "Epoch 5/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6788 - accuracy: 0.8067\n",
      "Epoch 00005: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 468ms/step - loss: 0.6788 - accuracy: 0.8067 - val_loss: 2.6592 - val_accuracy: 0.5800\n",
      "Epoch 6/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.6797 - accuracy: 0.8100\n",
      "Epoch 00006: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 494ms/step - loss: 0.6797 - accuracy: 0.8100 - val_loss: 3.6757 - val_accuracy: 0.4900\n",
      "Epoch 7/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3644 - accuracy: 0.8833\n",
      "Epoch 00007: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 481ms/step - loss: 0.3644 - accuracy: 0.8833 - val_loss: 3.1293 - val_accuracy: 0.5500\n",
      "Epoch 8/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2095 - accuracy: 0.9433\n",
      "Epoch 00008: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 481ms/step - loss: 0.2095 - accuracy: 0.9433 - val_loss: 2.8029 - val_accuracy: 0.5400\n",
      "Epoch 9/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1856 - accuracy: 0.9500\n",
      "Epoch 00009: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 475ms/step - loss: 0.1856 - accuracy: 0.9500 - val_loss: 2.7809 - val_accuracy: 0.5700\n",
      "Epoch 10/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2201 - accuracy: 0.9400\n",
      "Epoch 00010: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 495ms/step - loss: 0.2201 - accuracy: 0.9400 - val_loss: 3.6205 - val_accuracy: 0.5400\n",
      "Epoch 11/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2138 - accuracy: 0.9200\n",
      "Epoch 00011: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 478ms/step - loss: 0.2138 - accuracy: 0.9200 - val_loss: 2.9162 - val_accuracy: 0.5400\n",
      "Epoch 12/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1625 - accuracy: 0.9267\n",
      "Epoch 00012: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 467ms/step - loss: 0.1625 - accuracy: 0.9267 - val_loss: 2.6876 - val_accuracy: 0.6100\n",
      "Epoch 13/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1250 - accuracy: 0.9600\n",
      "Epoch 00013: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 465ms/step - loss: 0.1250 - accuracy: 0.9600 - val_loss: 2.5533 - val_accuracy: 0.5800\n",
      "Epoch 14/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1429 - accuracy: 0.9500\n",
      "Epoch 00014: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 472ms/step - loss: 0.1429 - accuracy: 0.9500 - val_loss: 2.8770 - val_accuracy: 0.5800\n",
      "Epoch 15/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1462 - accuracy: 0.9433\n",
      "Epoch 00015: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 482ms/step - loss: 0.1462 - accuracy: 0.9433 - val_loss: 2.8863 - val_accuracy: 0.5500\n",
      "Epoch 16/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0594 - accuracy: 0.9733\n",
      "Epoch 00016: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 479ms/step - loss: 0.0594 - accuracy: 0.9733 - val_loss: 2.7327 - val_accuracy: 0.5700\n",
      "Epoch 17/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1175 - accuracy: 0.9567\n",
      "Epoch 00017: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 482ms/step - loss: 0.1175 - accuracy: 0.9567 - val_loss: 2.5824 - val_accuracy: 0.6100\n",
      "Epoch 18/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0740 - accuracy: 0.9633\n",
      "Epoch 00018: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 492ms/step - loss: 0.0740 - accuracy: 0.9633 - val_loss: 2.7021 - val_accuracy: 0.5500\n",
      "Epoch 19/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1440 - accuracy: 0.9500\n",
      "Epoch 00019: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 469ms/step - loss: 0.1440 - accuracy: 0.9500 - val_loss: 3.2421 - val_accuracy: 0.6200\n",
      "Epoch 20/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2806 - accuracy: 0.9367\n",
      "Epoch 00020: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 478ms/step - loss: 0.2806 - accuracy: 0.9367 - val_loss: 3.3752 - val_accuracy: 0.5300\n",
      "Epoch 21/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2489 - accuracy: 0.9300\n",
      "Epoch 00021: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 458ms/step - loss: 0.2489 - accuracy: 0.9300 - val_loss: 4.3607 - val_accuracy: 0.4800\n",
      "Epoch 22/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.4326 - accuracy: 0.9000\n",
      "Epoch 00022: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 488ms/step - loss: 0.4326 - accuracy: 0.9000 - val_loss: 3.4234 - val_accuracy: 0.5100\n",
      "Epoch 23/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2428 - accuracy: 0.9300\n",
      "Epoch 00023: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 463ms/step - loss: 0.2428 - accuracy: 0.9300 - val_loss: 3.5330 - val_accuracy: 0.5600\n",
      "Epoch 24/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3083 - accuracy: 0.9100\n",
      "Epoch 00024: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 469ms/step - loss: 0.3083 - accuracy: 0.9100 - val_loss: 2.9254 - val_accuracy: 0.6000\n",
      "Epoch 25/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2078 - accuracy: 0.9433\n",
      "Epoch 00025: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 462ms/step - loss: 0.2078 - accuracy: 0.9433 - val_loss: 5.0115 - val_accuracy: 0.5300\n",
      "Epoch 26/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3140 - accuracy: 0.9233\n",
      "Epoch 00026: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 501ms/step - loss: 0.3140 - accuracy: 0.9233 - val_loss: 3.1433 - val_accuracy: 0.6100\n",
      "Epoch 27/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2312 - accuracy: 0.9267\n",
      "Epoch 00027: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 496ms/step - loss: 0.2312 - accuracy: 0.9267 - val_loss: 3.3646 - val_accuracy: 0.5600\n",
      "Epoch 28/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2725 - accuracy: 0.9500\n",
      "Epoch 00028: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 503ms/step - loss: 0.2725 - accuracy: 0.9500 - val_loss: 3.2224 - val_accuracy: 0.6100\n",
      "Epoch 29/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1775 - accuracy: 0.9500\n",
      "Epoch 00029: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 507ms/step - loss: 0.1775 - accuracy: 0.9500 - val_loss: 2.9827 - val_accuracy: 0.6000\n",
      "Epoch 30/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10/10 [==============================] - ETA: 0s - loss: 0.1125 - accuracy: 0.9633\n",
      "Epoch 00030: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 481ms/step - loss: 0.1125 - accuracy: 0.9633 - val_loss: 3.7087 - val_accuracy: 0.6000\n",
      "Epoch 31/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.3013 - accuracy: 0.9200\n",
      "Epoch 00031: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 508ms/step - loss: 0.3013 - accuracy: 0.9200 - val_loss: 3.8314 - val_accuracy: 0.6300\n",
      "Epoch 32/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.2645 - accuracy: 0.9500\n",
      "Epoch 00032: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 508ms/step - loss: 0.2645 - accuracy: 0.9500 - val_loss: 3.6607 - val_accuracy: 0.6100\n",
      "Epoch 33/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0981 - accuracy: 0.9767\n",
      "Epoch 00033: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 502ms/step - loss: 0.0981 - accuracy: 0.9767 - val_loss: 3.6512 - val_accuracy: 0.5900\n",
      "Epoch 34/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1527 - accuracy: 0.9567\n",
      "Epoch 00034: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 505ms/step - loss: 0.1527 - accuracy: 0.9567 - val_loss: 3.1159 - val_accuracy: 0.6000\n",
      "Epoch 35/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0900 - accuracy: 0.9800\n",
      "Epoch 00035: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 533ms/step - loss: 0.0900 - accuracy: 0.9800 - val_loss: 3.4292 - val_accuracy: 0.5900\n",
      "Epoch 36/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0942 - accuracy: 0.9700\n",
      "Epoch 00036: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.0942 - accuracy: 0.9700 - val_loss: 3.5230 - val_accuracy: 0.6000\n",
      "Epoch 37/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.1052 - accuracy: 0.9800\n",
      "Epoch 00037: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.1052 - accuracy: 0.9800 - val_loss: 3.3061 - val_accuracy: 0.6000\n",
      "Epoch 38/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0426 - accuracy: 0.9833\n",
      "Epoch 00038: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 475ms/step - loss: 0.0426 - accuracy: 0.9833 - val_loss: 3.0736 - val_accuracy: 0.5700\n",
      "Epoch 39/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0673 - accuracy: 0.9733\n",
      "Epoch 00039: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 490ms/step - loss: 0.0673 - accuracy: 0.9733 - val_loss: 3.1210 - val_accuracy: 0.6300\n",
      "Epoch 40/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0481 - accuracy: 0.9867\n",
      "Epoch 00040: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.0481 - accuracy: 0.9867 - val_loss: 3.1027 - val_accuracy: 0.6300\n",
      "Epoch 41/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0543 - accuracy: 0.9900\n",
      "Epoch 00041: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 460ms/step - loss: 0.0543 - accuracy: 0.9900 - val_loss: 3.2279 - val_accuracy: 0.6600\n",
      "Epoch 42/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0435 - accuracy: 0.9800\n",
      "Epoch 00042: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 476ms/step - loss: 0.0435 - accuracy: 0.9800 - val_loss: 3.1453 - val_accuracy: 0.6400\n",
      "Epoch 43/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0736 - accuracy: 0.9767\n",
      "Epoch 00043: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 468ms/step - loss: 0.0736 - accuracy: 0.9767 - val_loss: 3.3352 - val_accuracy: 0.5900\n",
      "Epoch 44/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0540 - accuracy: 0.9800\n",
      "Epoch 00044: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 484ms/step - loss: 0.0540 - accuracy: 0.9800 - val_loss: 3.5018 - val_accuracy: 0.5900\n",
      "Epoch 45/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0464 - accuracy: 0.9833\n",
      "Epoch 00045: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 472ms/step - loss: 0.0464 - accuracy: 0.9833 - val_loss: 3.6970 - val_accuracy: 0.5900\n",
      "Epoch 46/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0576 - accuracy: 0.9867\n",
      "Epoch 00046: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 475ms/step - loss: 0.0576 - accuracy: 0.9867 - val_loss: 4.2042 - val_accuracy: 0.5400\n",
      "Epoch 47/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0224 - accuracy: 0.9900\n",
      "Epoch 00047: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 452ms/step - loss: 0.0224 - accuracy: 0.9900 - val_loss: 3.5932 - val_accuracy: 0.5700\n",
      "Epoch 48/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0380 - accuracy: 0.9933\n",
      "Epoch 00048: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 451ms/step - loss: 0.0380 - accuracy: 0.9933 - val_loss: 3.6839 - val_accuracy: 0.5700\n",
      "Epoch 49/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0330 - accuracy: 0.9900\n",
      "Epoch 00049: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 461ms/step - loss: 0.0330 - accuracy: 0.9900 - val_loss: 3.8724 - val_accuracy: 0.6200\n",
      "Epoch 50/50\n",
      "10/10 [==============================] - ETA: 0s - loss: 0.0313 - accuracy: 0.9900\n",
      "Epoch 00050: saving model to best_xception_model.h5\n",
      "10/10 [==============================] - 5s 458ms/step - loss: 0.0313 - accuracy: 0.9900 - val_loss: 3.8864 - val_accuracy: 0.5600\n"
     ]
    }
   ],
   "source": [
    "r, model, test_set = create_model(Xception,'best_xception_model.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "366d50a8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 1s - loss: 3.8863 - accuracy: 0.5600 - 690ms/epoch - 173ms/step\n",
      "Restored model, accuracy: 56.00%\n"
     ]
    }
   ],
   "source": [
    "# Xception Model\n",
    "model.save('model_xception.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "77b460d8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4/4 - 1s - loss: 3.8863 - accuracy: 0.5600 - 1s/epoch - 333ms/step\n",
      "Restored model, accuracy: 56.00%\n"
     ]
    }
   ],
   "source": [
    "# Xception model while checkpointing\n",
    "model=load_model('best_xception_model.h5')\n",
    "loss, acc = model.evaluate(test_set, verbose=2)\n",
    "print(\"Restored model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "fcf422aa",
   "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": [
    "# plot the loss\n",
    "plt.plot(r.history['loss'], label='train loss')\n",
    "plt.plot(r.history['val_loss'], label='val loss')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('LossVal_loss')\n",
    "\n",
    "# plot the accuracy\n",
    "plt.plot(r.history['accuracy'], label='train acc')\n",
    "plt.plot(r.history['val_accuracy'], label='val acc')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "# plt.savefig('AccVal_acc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "06e29d9d",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "e3b09368",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.69592695e-07, 2.16649103e-04, 9.98911977e-01, 8.71018856e-04,\n",
       "        7.68050299e-08],\n",
       "       [2.08510357e-07, 2.99187164e-08, 4.85197660e-10, 1.19171684e-09,\n",
       "        9.99999762e-01],\n",
       "       [8.72781873e-02, 9.12067950e-01, 1.60937774e-11, 4.19266172e-07,\n",
       "        6.53482741e-04],\n",
       "       [1.91141769e-17, 5.17057820e-11, 1.40428210e-13, 1.00000000e+00,\n",
       "        1.11866676e-21],\n",
       "       [2.18767952e-03, 2.70300627e-01, 2.88959265e-01, 4.38549399e-01,\n",
       "        3.03914044e-06],\n",
       "       [2.20894050e-02, 9.64282095e-01, 5.86817798e-04, 2.12016417e-07,\n",
       "        1.30414730e-02],\n",
       "       [9.99999642e-01, 7.85722278e-08, 1.16247875e-07, 6.55460823e-08,\n",
       "        1.61555498e-08],\n",
       "       [1.81030780e-01, 1.24928564e-01, 4.00988611e-05, 8.56863509e-04,\n",
       "        6.93143725e-01],\n",
       "       [9.98506725e-01, 1.49323465e-03, 5.02720227e-11, 2.85895840e-09,\n",
       "        1.34879263e-09],\n",
       "       [5.72759609e-06, 3.35837007e-02, 9.66410339e-01, 8.71626611e-08,\n",
       "        1.66370555e-07],\n",
       "       [3.40615213e-01, 5.93777262e-02, 1.00592233e-01, 4.99414653e-01,\n",
       "        7.31779295e-08],\n",
       "       [1.84506553e-06, 5.39849736e-07, 1.04232676e-08, 2.38298998e-08,\n",
       "        9.99997616e-01],\n",
       "       [6.70307607e-04, 9.44349527e-01, 5.49678914e-02, 1.22376250e-05,\n",
       "        1.93730934e-08],\n",
       "       [1.16570227e-01, 8.83333921e-01, 3.20558228e-12, 5.36173866e-05,\n",
       "        4.22308731e-05],\n",
       "       [3.62604558e-01, 5.81608236e-01, 6.90554225e-05, 2.67807511e-04,\n",
       "        5.54503463e-02],\n",
       "       [1.07791518e-08, 3.40646948e-05, 2.34958225e-05, 9.99942422e-01,\n",
       "        9.52687151e-10],\n",
       "       [9.92642343e-01, 7.35771377e-03, 1.96507793e-11, 1.10770747e-13,\n",
       "        2.04012469e-08],\n",
       "       [8.04674089e-01, 1.79389268e-01, 1.57547463e-02, 3.22312953e-05,\n",
       "        1.49705957e-04],\n",
       "       [3.59801135e-07, 9.59754467e-01, 7.91517476e-15, 6.04206640e-09,\n",
       "        4.02452610e-02],\n",
       "       [3.02066510e-05, 1.80962343e-06, 2.48031830e-03, 9.97487664e-01,\n",
       "        1.50727182e-08],\n",
       "       [9.97775495e-01, 4.82632095e-05, 1.78342441e-03, 4.53564253e-09,\n",
       "        3.92775633e-04],\n",
       "       [1.03954971e-02, 4.75884910e-04, 9.84882176e-01, 1.77554059e-04,\n",
       "        4.06891666e-03],\n",
       "       [9.99615908e-01, 4.98892305e-09, 1.94124003e-16, 3.92942068e-10,\n",
       "        3.84039740e-04],\n",
       "       [4.90842410e-07, 4.52624577e-10, 7.57160634e-02, 9.24283445e-01,\n",
       "        1.52097779e-09],\n",
       "       [1.01870194e-03, 5.53664625e-01, 4.45316672e-01, 5.94870553e-09,\n",
       "        1.08668916e-11],\n",
       "       [6.29472866e-18, 8.22549140e-09, 6.88950008e-11, 1.00000000e+00,\n",
       "        4.33902622e-14],\n",
       "       [2.04730909e-02, 1.13391121e-04, 9.73506391e-01, 1.18481978e-04,\n",
       "        5.78860845e-03],\n",
       "       [1.11221686e-06, 6.22497191e-06, 9.99992132e-01, 4.32392568e-07,\n",
       "        2.96421145e-14],\n",
       "       [3.28147462e-06, 4.01932746e-04, 4.42504026e-02, 9.55344439e-01,\n",
       "        1.78762595e-11],\n",
       "       [6.97799385e-01, 3.00109029e-01, 5.21324750e-07, 8.29908313e-05,\n",
       "        2.00802879e-03],\n",
       "       [1.36180400e-08, 2.77518293e-08, 1.45264005e-06, 5.13860881e-01,\n",
       "        4.86137569e-01],\n",
       "       [1.76920403e-05, 9.94165778e-01, 9.78836370e-13, 2.15020407e-07,\n",
       "        5.81624499e-03],\n",
       "       [7.89181329e-04, 3.53971473e-03, 9.10142720e-01, 8.01640842e-03,\n",
       "        7.75118694e-02],\n",
       "       [6.24712805e-17, 3.28784076e-08, 2.43445726e-13, 1.00000000e+00,\n",
       "        5.42863783e-23],\n",
       "       [2.19211923e-11, 4.17558960e-10, 9.30860937e-01, 6.91390783e-02,\n",
       "        9.56121900e-12],\n",
       "       [1.47291915e-02, 3.34196142e-03, 9.81801450e-01, 1.27274165e-04,\n",
       "        6.64294930e-08],\n",
       "       [3.93029859e-06, 5.41273812e-06, 9.99990702e-01, 4.29499742e-08,\n",
       "        5.22260533e-11],\n",
       "       [2.00979677e-09, 1.14101104e-10, 1.00000000e+00, 2.31745925e-11,\n",
       "        1.12517017e-13],\n",
       "       [1.21023199e-02, 5.70544064e-01, 3.37930808e-14, 5.85067314e-13,\n",
       "        4.17353660e-01],\n",
       "       [1.11341395e-03, 1.89705506e-01, 2.09063122e-09, 2.68791041e-07,\n",
       "        8.09180856e-01],\n",
       "       [3.02490545e-04, 5.52856887e-04, 5.87103716e-12, 8.64910921e-11,\n",
       "        9.99144673e-01],\n",
       "       [3.13550569e-02, 9.67687845e-01, 9.31964838e-04, 3.46277318e-08,\n",
       "        2.52014524e-05],\n",
       "       [6.66075503e-05, 5.17127373e-05, 3.32290064e-11, 4.36958887e-14,\n",
       "        9.99881625e-01],\n",
       "       [5.71196433e-03, 9.93884027e-01, 5.67263378e-05, 3.46951856e-04,\n",
       "        3.12791059e-07],\n",
       "       [1.58742496e-05, 3.03109264e-04, 2.36108858e-11, 4.12626340e-13,\n",
       "        9.99680996e-01],\n",
       "       [2.91783499e-06, 9.99890566e-01, 1.49153720e-13, 1.21828060e-11,\n",
       "        1.06568201e-04],\n",
       "       [9.96284962e-01, 1.80198229e-03, 1.05716085e-08, 1.33026106e-07,\n",
       "        1.91301003e-03],\n",
       "       [1.66686152e-06, 2.95320788e-05, 2.58892530e-09, 9.35173361e-09,\n",
       "        9.99968767e-01],\n",
       "       [1.00000000e+00, 8.46210435e-10, 1.71767505e-19, 3.09741348e-19,\n",
       "        1.90830352e-09],\n",
       "       [4.57262397e-02, 9.54028308e-01, 1.68469292e-08, 2.45449366e-04,\n",
       "        4.97519304e-10],\n",
       "       [3.24819255e-10, 1.20477651e-07, 9.99332011e-01, 2.49351084e-04,\n",
       "        4.18471725e-04],\n",
       "       [2.11198494e-04, 2.08441764e-02, 1.84393976e-11, 3.79714066e-10,\n",
       "        9.78944659e-01],\n",
       "       [8.75134447e-06, 3.19132693e-02, 8.14359188e-01, 1.31279439e-01,\n",
       "        2.24393569e-02],\n",
       "       [1.22868805e-05, 8.92655225e-06, 9.97369170e-01, 2.60968530e-03,\n",
       "        1.18014010e-09],\n",
       "       [4.51202097e-04, 9.98035371e-01, 1.59720555e-12, 6.32840932e-11,\n",
       "        1.51345716e-03],\n",
       "       [9.33125854e-01, 6.24204502e-02, 2.84915119e-10, 1.52715046e-07,\n",
       "        4.45357710e-03],\n",
       "       [4.72055175e-13, 2.22431084e-11, 3.61294538e-01, 6.38705432e-01,\n",
       "        1.61984847e-16],\n",
       "       [4.05902475e-01, 3.12024981e-01, 1.16032808e-10, 3.12964630e-06,\n",
       "        2.82069385e-01],\n",
       "       [1.83809902e-07, 3.27985764e-01, 1.10676706e-01, 2.30065703e-10,\n",
       "        5.61337352e-01],\n",
       "       [5.14755186e-07, 5.16588261e-06, 9.99994040e-01, 2.20415302e-07,\n",
       "        4.95596169e-08],\n",
       "       [1.74895192e-08, 9.99931097e-01, 1.12862308e-05, 3.83163933e-06,\n",
       "        5.37918131e-05],\n",
       "       [7.01960300e-13, 3.93069639e-11, 2.75642009e-12, 1.00000000e+00,\n",
       "        2.20757267e-14],\n",
       "       [1.00000000e+00, 3.82613330e-10, 2.24191085e-20, 1.16023360e-21,\n",
       "        7.34846298e-12],\n",
       "       [6.44666276e-09, 1.01498088e-07, 9.99999762e-01, 1.14546737e-07,\n",
       "        5.94695395e-14],\n",
       "       [1.02726131e-04, 2.64105620e-04, 1.86629314e-02, 9.80967820e-01,\n",
       "        2.47652019e-06],\n",
       "       [9.99426126e-01, 1.14994118e-05, 6.46485205e-05, 1.25384622e-05,\n",
       "        4.85175668e-04],\n",
       "       [3.10623705e-01, 6.89376235e-01, 8.23645430e-10, 9.90487266e-08,\n",
       "        6.99041790e-13],\n",
       "       [6.55812837e-05, 2.03023064e-05, 1.61679596e-01, 8.38234544e-01,\n",
       "        6.21140223e-11],\n",
       "       [6.59587658e-06, 6.83906257e-01, 8.29554945e-02, 2.29922757e-01,\n",
       "        3.20893642e-03],\n",
       "       [3.31694428e-05, 2.22893621e-04, 2.73708682e-12, 7.77089681e-08,\n",
       "        9.99743879e-01],\n",
       "       [1.04158567e-02, 9.89488900e-01, 2.73718648e-09, 9.26716020e-05,\n",
       "        2.51912797e-06],\n",
       "       [1.22809699e-02, 2.19149500e-04, 5.17612338e-01, 4.69887137e-01,\n",
       "        4.27757442e-07],\n",
       "       [9.99734342e-01, 2.50770012e-04, 1.48651370e-05, 3.75668341e-09,\n",
       "        9.21457702e-11],\n",
       "       [9.49678824e-09, 5.21456423e-06, 1.23493909e-03, 9.98759866e-01,\n",
       "        6.84302037e-09],\n",
       "       [1.19343415e-08, 2.65922234e-03, 8.54190469e-01, 1.43149808e-01,\n",
       "        4.99503642e-07],\n",
       "       [5.41240865e-07, 1.35736480e-01, 9.42699808e-06, 8.64253342e-01,\n",
       "        1.43861115e-07],\n",
       "       [8.50578090e-06, 8.57350187e-07, 9.99990582e-01, 6.70693678e-09,\n",
       "        9.69435435e-11],\n",
       "       [8.84164155e-01, 4.51538572e-03, 7.30174258e-02, 3.83030511e-02,\n",
       "        6.85366430e-10],\n",
       "       [9.75427926e-01, 1.75242294e-02, 7.15103988e-06, 6.58136979e-03,\n",
       "        4.59252566e-04],\n",
       "       [9.98634875e-01, 1.36511330e-03, 9.10640400e-12, 4.26322241e-11,\n",
       "        2.03693773e-09],\n",
       "       [1.22042565e-09, 1.45550803e-05, 3.80387610e-05, 9.99947429e-01,\n",
       "        2.61759218e-14],\n",
       "       [3.03978431e-05, 9.03393084e-05, 1.60079549e-18, 6.47297754e-13,\n",
       "        9.99879241e-01],\n",
       "       [1.75836778e-09, 5.09741120e-08, 9.96692777e-01, 3.29690566e-03,\n",
       "        1.03309203e-05],\n",
       "       [6.93080153e-08, 2.49890201e-02, 4.36962087e-04, 9.74564314e-01,\n",
       "        9.65437084e-06],\n",
       "       [9.97773707e-01, 2.22625746e-03, 3.93117787e-12, 2.08992147e-11,\n",
       "        1.29428717e-08],\n",
       "       [4.33607548e-01, 5.66291809e-01, 1.26121086e-12, 1.10217347e-11,\n",
       "        1.00663805e-04],\n",
       "       [2.60877073e-06, 8.63172787e-08, 9.97264624e-01, 2.73267156e-03,\n",
       "        1.02130811e-14],\n",
       "       [9.25346394e-05, 8.21239496e-07, 9.99890447e-01, 1.61756398e-05,\n",
       "        8.58752291e-10],\n",
       "       [2.15158681e-03, 9.42131102e-01, 5.36108166e-02, 2.10659578e-03,\n",
       "        1.04254543e-08],\n",
       "       [2.78091562e-07, 1.78725561e-04, 9.99265492e-01, 5.55147766e-04,\n",
       "        3.88727244e-07],\n",
       "       [2.84467892e-12, 1.18345611e-10, 4.08169217e-02, 9.59183156e-01,\n",
       "        5.99821538e-12],\n",
       "       [9.99579728e-01, 4.19002201e-04, 3.87173713e-12, 2.90028522e-11,\n",
       "        1.25132874e-06],\n",
       "       [6.23937346e-10, 1.00000000e+00, 4.92740224e-12, 3.95646227e-09,\n",
       "        4.18331680e-08],\n",
       "       [6.54429436e-01, 3.45451742e-01, 4.62666549e-07, 2.40470254e-05,\n",
       "        9.43082268e-05],\n",
       "       [7.35243876e-03, 7.05110192e-01, 1.50619712e-06, 2.87535042e-01,\n",
       "        7.65875313e-07],\n",
       "       [8.04268041e-10, 1.43872668e-07, 9.99999166e-01, 6.68467635e-07,\n",
       "        2.61771263e-14],\n",
       "       [8.77649953e-09, 1.40688999e-05, 3.69392739e-10, 1.08988096e-11,\n",
       "        9.99985933e-01],\n",
       "       [1.77117521e-09, 9.67691064e-01, 8.31431771e-06, 3.23005691e-02,\n",
       "        2.31142501e-11],\n",
       "       [1.32080795e-05, 1.24836876e-03, 9.98738348e-01, 6.86347761e-08,\n",
       "        1.19651915e-08],\n",
       "       [5.66934049e-02, 9.25238788e-01, 9.17925427e-05, 3.20134894e-03,\n",
       "        1.47747714e-02]], dtype=float32)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict(test_set)\n",
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "46107bc0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2, 4, 1, 3, 3, 1, 0, 4, 0, 2, 3, 4, 1, 1, 1, 3, 0, 0, 1, 3, 0, 2,\n",
       "       0, 3, 1, 3, 2, 2, 3, 0, 3, 1, 2, 3, 2, 2, 2, 2, 1, 4, 4, 1, 4, 1,\n",
       "       4, 1, 0, 4, 0, 1, 2, 4, 2, 2, 1, 0, 3, 0, 4, 2, 1, 3, 0, 2, 3, 0,\n",
       "       1, 3, 1, 4, 1, 2, 0, 3, 2, 3, 2, 0, 0, 0, 3, 4, 2, 3, 0, 1, 2, 2,\n",
       "       1, 2, 3, 0, 1, 0, 1, 2, 4, 1, 2, 1], dtype=int64)"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = numpy.argmax(y_pred,axis=1)\n",
    "y_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "bf8aeb65",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        ...,\n",
       "        [212., 205., 199.],\n",
       "        [207., 207., 199.],\n",
       "        [207., 207., 199.]],\n",
       "\n",
       "       [[199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        ...,\n",
       "        [212., 205., 199.],\n",
       "        [207., 207., 199.],\n",
       "        [207., 207., 199.]],\n",
       "\n",
       "       [[199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        [199., 199., 197.],\n",
       "        ...,\n",
       "        [211., 204., 198.],\n",
       "        [206., 206., 198.],\n",
       "        [206., 206., 198.]],\n",
       "\n",
       "       ...,\n",
       "\n",
       "       [[151., 104.,  62.],\n",
       "        [143.,  96.,  54.],\n",
       "        [153., 108.,  66.],\n",
       "        ...,\n",
       "        [143.,  98.,  56.],\n",
       "        [140.,  91.,  48.],\n",
       "        [141.,  92.,  49.]],\n",
       "\n",
       "       [[150., 103.,  61.],\n",
       "        [144.,  97.,  55.],\n",
       "        [146., 101.,  60.],\n",
       "        ...,\n",
       "        [137.,  92.,  50.],\n",
       "        [146.,  97.,  54.],\n",
       "        [141.,  94.,  50.]],\n",
       "\n",
       "       [[146.,  99.,  57.],\n",
       "        [141.,  93.,  53.],\n",
       "        [134.,  89.,  48.],\n",
       "        ...,\n",
       "        [168., 123.,  81.],\n",
       "        [143.,  96.,  52.],\n",
       "        [134.,  87.,  43.]]], dtype=float32)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img = image.load_img('Datasets/Test/bmw/7.jpg',target_size=(224,224))\n",
    "x = image.img_to_array(img)\n",
    "x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "6c7748db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(224, 224, 3)"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "a53beffa",
   "metadata": {},
   "outputs": [],
   "source": [
    "x = x/255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "282f919e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1, 224, 224, 3)"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x = numpy.expand_dims(x, axis=0)\n",
    "img_data = preprocess_input(x)\n",
    "img_data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "e453dfaa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.0000000e+00, 9.9968088e-01, 0.0000000e+00, 3.1908552e-04,\n",
       "        0.0000000e+00]], dtype=float32)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred = model.predict(img_data)\n",
    "pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "c478be52",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred = numpy.argmax(pred)\n",
    "pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "e9295f10",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Car brand is Bmw\n"
     ]
    }
   ],
   "source": [
    "print('Car brand is',os.listdir(test_path)[pred].capitalize())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4862f9d5",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}