PDD_Classification_Niranjan.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 528989,
     "status": "ok",
     "timestamp": 1623172892175,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "pBUPYa6yQbVc",
    "outputId": "09742ccf-a326-47c5-919f-7b17e619285d"
   },
   "outputs": [],
   "source": [
    "# # Download a file based on its file ID.\r\n",
    "# file_id = '18DbC6Xj4NP-hLzI14WuMaAEyq482vNfn'\r\n",
    "\r\n",
    "# # Download dataset\r\n",
    "# !gdown https://drive.google.com/uc?id={file_id}\r\n",
    "\r\n",
    "# # Unzip the downloaded file\r\n",
    "# !unzip -q PlantVillage.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "executionInfo": {
     "elapsed": 9244,
     "status": "ok",
     "timestamp": 1623177820513,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "ckkfHCP7rJhi"
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential, Model\r\n",
    "from tensorflow.keras.layers import Conv2D,Activation,LeakyReLU,BatchNormalization,MaxPooling2D,Flatten,Dense,Dropout\r\n",
    "from tensorflow.keras.optimizers import Adam\r\n",
    "from tensorflow.keras import backend as K\r\n",
    "\r\n",
    "\r\n",
    "from tensorflow.keras.preprocessing import image\r\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator,img_to_array\r\n",
    "import xgboost as xgb\r\n",
    "import matplotlib.pyplot as plt\r\n",
    "import pandas as pd\r\n",
    "import numpy as np\r\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "executionInfo": {
     "elapsed": 654,
     "status": "ok",
     "timestamp": 1623177821157,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "0Zqy61__Rn5L"
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "import cv2\n",
    "import os\n",
    "from os import listdir\n",
    "from sklearn.preprocessing import LabelBinarizer,MultiLabelBinarizer\n",
    "from sklearn.model_selection import train_test_split\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['train', 'val']"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "root_dir = r'C:\\Users\\niran\\Desktop\\PDD_Algorithm\\PlantVillage'\r\n",
    "\r\n",
    "os.chdir(root_dir)\r\n",
    "listdir()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "executionInfo": {
     "elapsed": 371,
     "status": "ok",
     "timestamp": 1623177823254,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "GZ93_IycTZPa"
   },
   "outputs": [],
   "source": [
    "# Dimension of resized image\r\n",
    "DEFAULT_IMAGE_SIZE = tuple((256, 256))\r\n",
    "\r\n",
    "# Number of images used to train the model\r\n",
    "N_IMAGES = 100\r\n",
    "\r\n",
    "data_dir = os.path.join(root_dir,'train')\r\n",
    "\r\n",
    "\"\"\"We use the function `convert_image_to_array` to resize an image to the size `DEFAULT_IMAGE_SIZE` we defined above.\"\"\"\r\n",
    "\r\n",
    "def convert_image_to_array(image_dir):\r\n",
    "    try:\r\n",
    "        image = cv2.imread(image_dir)\r\n",
    "        if image is not None:\r\n",
    "            image = cv2.resize(image, DEFAULT_IMAGE_SIZE)   \r\n",
    "            return img_to_array(image)\r\n",
    "        else:\r\n",
    "            return np.array([])\r\n",
    "    except Exception as e:\r\n",
    "        print(f\"Error : {e}\")\r\n",
    "        return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11,
     "status": "ok",
     "timestamp": 1623177825147,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "QIXPhJmoTvyv",
    "outputId": "ba34dd43-b38c-4d19-dddc-40a942cc26e5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Load images from all classes ...\n",
      "39\n"
     ]
    }
   ],
   "source": [
    "print(\"Load images from all classes ...\")\r\n",
    "plant_disease_folder_list = listdir(data_dir)\r\n",
    "print(len(plant_disease_folder_list))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Loading Training & Validation Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 10954,
     "status": "ok",
     "timestamp": 1623177838050,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "VgUTu7ZFTjaD",
    "outputId": "9fc82226-e80e-4e72-f852-86fab68e327e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loding Image Data ...\n",
      "Processing  Apple___Apple_scab ...\n",
      "Processing  Apple___Black_rot ...\n",
      "Processing  Apple___Cedar_apple_rust ...\n",
      "Processing  Apple___healthy ...\n",
      "Processing  background ...\n",
      "Processing  Blueberry___healthy ...\n",
      "Processing  Cherry_(including_sour)___healthy ...\n",
      "Processing  Cherry_(including_sour)___Powdery_mildew ...\n",
      "Processing  Corn_(maize)___Cercospora_leaf_spot Gray_leaf_spot ...\n",
      "Processing  Corn_(maize)___Common_rust_ ...\n",
      "Processing  Corn_(maize)___healthy ...\n",
      "Processing  Corn_(maize)___Northern_Leaf_Blight ...\n",
      "Processing  Grape___Black_rot ...\n",
      "Processing  Grape___Esca_(Black_Measles) ...\n",
      "Processing  Grape___healthy ...\n",
      "Processing  Grape___Leaf_blight_(Isariopsis_Leaf_Spot) ...\n",
      "Processing  Orange___Haunglongbing_(Citrus_greening) ...\n",
      "Processing  Peach___Bacterial_spot ...\n",
      "Processing  Peach___healthy ...\n",
      "Processing  Pepper,_bell___Bacterial_spot ...\n",
      "Processing  Pepper,_bell___healthy ...\n",
      "Processing  Potato___Early_blight ...\n",
      "Processing  Potato___healthy ...\n",
      "Processing  Potato___Late_blight ...\n",
      "Processing  Raspberry___healthy ...\n",
      "Processing  Soybean___healthy ...\n",
      "Processing  Squash___Powdery_mildew ...\n",
      "Processing  Strawberry___healthy ...\n",
      "Processing  Strawberry___Leaf_scorch ...\n",
      "Processing  Tomato___Bacterial_spot ...\n",
      "Processing  Tomato___Early_blight ...\n",
      "Processing  Tomato___healthy ...\n",
      "Processing  Tomato___Late_blight ...\n",
      "Processing  Tomato___Leaf_Mold ...\n",
      "Processing  Tomato___Septoria_leaf_spot ...\n",
      "Processing  Tomato___Spider_mites Two-spotted_spider_mite ...\n",
      "Processing  Tomato___Target_Spot ...\n",
      "Processing  Tomato___Tomato_mosaic_virus ...\n",
      "Processing  Tomato___Tomato_Yellow_Leaf_Curl_Virus ...\n",
      "Processing  Apple___Apple_scab ...\n",
      "Processing  Apple___Black_rot ...\n",
      "Processing  Apple___Cedar_apple_rust ...\n",
      "Processing  Apple___healthy ...\n",
      "Processing  background ...\n",
      "Processing  Blueberry___healthy ...\n",
      "Processing  Cherry_(including_sour)___healthy ...\n",
      "Processing  Cherry_(including_sour)___Powdery_mildew ...\n",
      "Processing  Corn_(maize)___Cercospora_leaf_spot Gray_leaf_spot ...\n",
      "Processing  Corn_(maize)___Common_rust_ ...\n",
      "Processing  Corn_(maize)___healthy ...\n",
      "Processing  Corn_(maize)___Northern_Leaf_Blight ...\n",
      "Processing  Grape___Black_rot ...\n",
      "Processing  Grape___Esca_(Black_Measles) ...\n",
      "Processing  Grape___healthy ...\n",
      "Processing  Grape___Leaf_blight_(Isariopsis_Leaf_Spot) ...\n",
      "Processing  Orange___Haunglongbing_(Citrus_greening) ...\n",
      "Processing  Peach___Bacterial_spot ...\n",
      "Processing  Peach___healthy ...\n",
      "Processing  Pepper,_bell___Bacterial_spot ...\n",
      "Processing  Pepper,_bell___healthy ...\n",
      "Processing  Potato___Early_blight ...\n",
      "Processing  Potato___healthy ...\n",
      "Processing  Potato___Late_blight ...\n",
      "Processing  Raspberry___healthy ...\n",
      "Processing  Soybean___healthy ...\n",
      "Processing  Squash___Powdery_mildew ...\n",
      "Processing  Strawberry___healthy ...\n",
      "Processing  Strawberry___Leaf_scorch ...\n",
      "Processing  Tomato___Bacterial_spot ...\n",
      "Processing  Tomato___Early_blight ...\n",
      "Processing  Tomato___healthy ...\n",
      "Processing  Tomato___Late_blight ...\n",
      "Processing  Tomato___Leaf_Mold ...\n",
      "Processing  Tomato___Septoria_leaf_spot ...\n",
      "Processing  Tomato___Spider_mites Two-spotted_spider_mite ...\n",
      "Processing  Tomato___Target_Spot ...\n",
      "Processing  Tomato___Tomato_mosaic_virus ...\n",
      "Processing  Tomato___Tomato_Yellow_Leaf_Curl_Virus ...\n",
      "All the images have successfully loaded!!\n"
     ]
    }
   ],
   "source": [
    "image_list, label_list = [], []\r\n",
    "\r\n",
    "try:\r\n",
    "    print(\"Loding Image Data ...\")\r\n",
    "    for s in listdir():\r\n",
    "        plant_disease_folder_list = listdir(s)\r\n",
    "\r\n",
    "        for plant_disease_folder in plant_disease_folder_list:\r\n",
    "            print(f\"Processing  {plant_disease_folder} ...\")\r\n",
    "            plant_disease_image_list = listdir(f\"{data_dir}/{plant_disease_folder}/\")\r\n",
    "\r\n",
    "            for image in plant_disease_image_list[:N_IMAGES]:\r\n",
    "                image_directory = f\"{data_dir}/{plant_disease_folder}/{image}\"\r\n",
    "                if image_directory.endswith(\".jpg\")==True or image_directory.endswith(\".JPG\")==True:\r\n",
    "                    image_list.append(convert_image_to_array(image_directory))\r\n",
    "                    label_list.append(plant_disease_folder)\r\n",
    "\r\n",
    "    print(\"All the images have successfully loaded!!\")  \r\n",
    "except Exception as e:\r\n",
    "    print(f\"Error : {e}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1623177838050,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "Cu2DGtVGW3r3",
    "outputId": "aebba22f-7a88-4210-88f1-43597cc48156"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7800\n"
     ]
    }
   ],
   "source": [
    "print(len(image_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 11984,
     "status": "ok",
     "timestamp": 1623177852520,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "AgXTKzClTjLY",
    "outputId": "f9aafeb2-f6c1-477f-dbfb-2a0dd52bf1be"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of images: 7800\n"
     ]
    }
   ],
   "source": [
    "# Transform the loaded training image data into numpy array\r\n",
    "np_image_list = np.array(image_list, dtype=np.float16) / 255.0\r\n",
    "\r\n",
    "# Check the number of images loaded for training\r\n",
    "image_len = len(image_list)\r\n",
    "print(f\"Total number of images: {image_len}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# One-Hot Encoding the labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 16,
     "status": "ok",
     "timestamp": 1623177852520,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "15EUEA6dUjAb",
    "outputId": "60702bd3-a1bb-4940-a20e-7fd42d9cdcd7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of classes:  39\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Examine the labels/classes in the training dataset.\"\"\"\r\n",
    "\r\n",
    "label_binarizer = LabelBinarizer()\r\n",
    "image_labels = label_binarizer.fit_transform(label_list)\r\n",
    "\r\n",
    "# pickle.dump(label_binarizer,open('plant_disease_label_transform.pkl', 'wb'))\r\n",
    "n_classes = len(label_binarizer.classes_)\r\n",
    "\r\n",
    "print(\"Total number of classes: \", n_classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "executionInfo": {
     "elapsed": 6,
     "status": "ok",
     "timestamp": 1623177852521,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "aofjXWyxTjCq"
   },
   "outputs": [],
   "source": [
    "\"\"\"# Augment and Split Dataset\n",
    "Using `ImageDataGenerator` to augment data by performing various operations on the training images.\n",
    "\"\"\"\n",
    "\n",
    "augment = ImageDataGenerator(rotation_range=25, width_shift_range=0.1,\n",
    "                             height_shift_range=0.1, shear_range=0.2, \n",
    "                             zoom_range=0.2, horizontal_flip=True, \n",
    "                             fill_mode=\"nearest\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 625,
     "status": "ok",
     "timestamp": 1623177853140,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "lQglwnzGYPA5",
    "outputId": "97d37b61-bf7a-4015-e0ea-5561fe1c54ab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Successfully split data into TRAIN & TEST\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Splitting the data into training and test sets for validation purpose.\"\"\"\r\n",
    "\r\n",
    "x_train, x_test, y_train, y_test = train_test_split(np_image_list, image_labels, test_size=0.2, random_state = 42)\r\n",
    "print('Successfully split data into TRAIN & TEST')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 6045,
     "status": "ok",
     "timestamp": 1623177873583,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "a6CiKlFzrIV2",
    "outputId": "e950556d-0b24-419e-a187-72af400f6884"
   },
   "outputs": [],
   "source": [
    "\"\"\"# Build Model\r\n",
    "Defining the hyperparameters of the plant disease classification model.\r\n",
    "\"\"\"\r\n",
    "\r\n",
    "EPOCHS = 10\r\n",
    "STEPS = 100\r\n",
    "LR = 1e-3\r\n",
    "BATCH_SIZE = 32\r\n",
    "WIDTH = 256\r\n",
    "HEIGHT = 256\r\n",
    "DEPTH = 3\r\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 256, 256, 32)      896       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu (LeakyReLU)      (None, 256, 256, 32)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 256, 256, 32)      128       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 85, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 85, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 85, 85, 64)        18496     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_1 (LeakyReLU)    (None, 85, 85, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 85, 85, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 85, 85, 64)        36928     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_2 (LeakyReLU)    (None, 85, 85, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 85, 85, 64)        256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 42, 42, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 42, 42, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 42, 42, 128)       73856     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_3 (LeakyReLU)    (None, 42, 42, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 42, 42, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 42, 42, 128)       147584    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_4 (LeakyReLU)    (None, 42, 42, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 42, 42, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 21, 21, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 21, 21, 128)       0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 56448)             0         \n",
      "_________________________________________________________________\n",
      "my_dense (Dense)             (None, 1024)              57803776  \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_5 (LeakyReLU)    (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 1024)              4096      \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 39)                39975     \n",
      "_________________________________________________________________\n",
      "activation (Activation)      (None, 39)                0         \n",
      "=================================================================\n",
      "Total params: 58,127,271\n",
      "Trainable params: 58,124,391\n",
      "Non-trainable params: 2,880\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Creating a Sequential Model to build CNN for multi-class classification\"\"\"\r\n",
    "\r\n",
    "model = Sequential()\r\n",
    "inputShape = (HEIGHT, WIDTH, DEPTH)\r\n",
    "chanDim = -1\r\n",
    "\r\n",
    "if K.image_data_format() == \"channels_first\":\r\n",
    "    inputShape = (DEPTH, HEIGHT, WIDTH)\r\n",
    "    chanDim = 1\r\n",
    "\r\n",
    "model.add(Conv2D(32, (3, 3), padding=\"same\",input_shape=inputShape))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization(axis=chanDim))\r\n",
    "model.add(MaxPooling2D(pool_size=(3, 3)))\r\n",
    "model.add(Dropout(0.25))\r\n",
    "\r\n",
    "model.add(Conv2D(64, (3, 3), padding=\"same\"))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization(axis=chanDim))\r\n",
    "\r\n",
    "model.add(Conv2D(64, (3, 3), padding=\"same\"))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization(axis=chanDim))\r\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\r\n",
    "model.add(Dropout(0.25))\r\n",
    "\r\n",
    "model.add(Conv2D(128, (3, 3), padding=\"same\"))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization(axis=chanDim))\r\n",
    "\r\n",
    "model.add(Conv2D(128, (3, 3), padding=\"same\"))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization(axis=chanDim))\r\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\r\n",
    "model.add(Dropout(0.25))\r\n",
    "\r\n",
    "model.add(Flatten())\r\n",
    "\r\n",
    "model.add(Dense(1024, name = 'my_dense'))\r\n",
    "model.add(LeakyReLU(alpha=0.1))\r\n",
    "model.add(BatchNormalization())\r\n",
    "model.add(Dropout(0.5))\r\n",
    "\r\n",
    "model.add(Dense(n_classes))\r\n",
    "model.add(Activation(\"softmax\"))\r\n",
    "\r\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Extraction from the Dense Layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 458,
     "status": "ok",
     "timestamp": 1623177880481,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "lrPzn71H6Wdp",
    "outputId": "f5c28751-cf41-445c-b834-45aac23d36e4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_input (InputLayer)    [(None, 256, 256, 3)]     0         \n",
      "_________________________________________________________________\n",
      "conv2d (Conv2D)              (None, 256, 256, 32)      896       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu (LeakyReLU)      (None, 256, 256, 32)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization (BatchNo (None, 256, 256, 32)      128       \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 85, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 85, 85, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 85, 85, 64)        18496     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_1 (LeakyReLU)    (None, 85, 85, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 85, 85, 64)        256       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 85, 85, 64)        36928     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_2 (LeakyReLU)    (None, 85, 85, 64)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 85, 85, 64)        256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 42, 42, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 42, 42, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 42, 42, 128)       73856     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_3 (LeakyReLU)    (None, 42, 42, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 42, 42, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 42, 42, 128)       147584    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_4 (LeakyReLU)    (None, 42, 42, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 42, 42, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 21, 21, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 21, 21, 128)       0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 56448)             0         \n",
      "_________________________________________________________________\n",
      "my_dense (Dense)             (None, 1024)              57803776  \n",
      "=================================================================\n",
      "Total params: 58,083,200\n",
      "Trainable params: 58,082,368\n",
      "Non-trainable params: 832\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "layer_name = 'my_dense'\r\n",
    "intermediate_layer_model = Model(inputs=model.input,outputs=model.get_layer(layer_name).output) \r\n",
    "intermediate_layer_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "executionInfo": {
     "elapsed": 826220,
     "status": "ok",
     "timestamp": 1623178709653,
     "user": {
      "displayName": "Akella Niranjan",
      "photoUrl": "https://lh3.googleusercontent.com/a-/AOh14GiS0hDW7cYrcLRQ7IP_TVEsKTd1iwuhmSjd-QBh=s64",
      "userId": "10548223611542786095"
     },
     "user_tz": -330
    },
    "id": "5FGRh6ojt96j",
    "outputId": "4cfa34e8-8760-4cca-e196-f02394234286"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training CNN...\n",
      "Epoch 1/20\n",
      "195/195 [==============================] - 108s 379ms/step - loss: 0.4214 - accuracy: 0.1059 - val_loss: 0.1297 - val_accuracy: 0.0276\n",
      "Epoch 2/20\n",
      "195/195 [==============================] - 76s 388ms/step - loss: 0.0951 - accuracy: 0.4194 - val_loss: 0.1158 - val_accuracy: 0.1891\n",
      "Epoch 3/20\n",
      "195/195 [==============================] - 74s 380ms/step - loss: 0.0703 - accuracy: 0.5747 - val_loss: 0.0823 - val_accuracy: 0.4122\n",
      "Epoch 4/20\n",
      "195/195 [==============================] - 75s 385ms/step - loss: 0.0593 - accuracy: 0.6404 - val_loss: 0.0705 - val_accuracy: 0.5724\n",
      "Epoch 5/20\n",
      "195/195 [==============================] - 74s 380ms/step - loss: 0.0504 - accuracy: 0.6998 - val_loss: 0.0795 - val_accuracy: 0.5321\n",
      "Epoch 6/20\n",
      "195/195 [==============================] - 74s 377ms/step - loss: 0.0437 - accuracy: 0.7441 - val_loss: 0.0766 - val_accuracy: 0.5827\n",
      "Epoch 7/20\n",
      "195/195 [==============================] - 73s 374ms/step - loss: 0.0388 - accuracy: 0.7753 - val_loss: 0.0814 - val_accuracy: 0.5660\n",
      "Epoch 8/20\n",
      "195/195 [==============================] - 72s 369ms/step - loss: 0.0338 - accuracy: 0.8074 - val_loss: 0.0745 - val_accuracy: 0.5936\n",
      "Epoch 9/20\n",
      "195/195 [==============================] - 71s 362ms/step - loss: 0.0315 - accuracy: 0.8277 - val_loss: 0.0457 - val_accuracy: 0.7359\n",
      "Epoch 10/20\n",
      "195/195 [==============================] - 72s 366ms/step - loss: 0.0295 - accuracy: 0.8401 - val_loss: 0.0849 - val_accuracy: 0.5481\n",
      "Epoch 11/20\n",
      "195/195 [==============================] - 70s 360ms/step - loss: 0.0265 - accuracy: 0.8614 - val_loss: 0.0474 - val_accuracy: 0.7109\n",
      "Epoch 12/20\n",
      "195/195 [==============================] - 70s 359ms/step - loss: 0.0240 - accuracy: 0.8740 - val_loss: 0.0442 - val_accuracy: 0.7526\n",
      "Epoch 13/20\n",
      "195/195 [==============================] - 70s 356ms/step - loss: 0.0226 - accuracy: 0.8902 - val_loss: 0.0259 - val_accuracy: 0.8660\n",
      "Epoch 14/20\n",
      "195/195 [==============================] - 70s 357ms/step - loss: 0.0214 - accuracy: 0.8920 - val_loss: 0.1363 - val_accuracy: 0.4596\n",
      "Epoch 15/20\n",
      "195/195 [==============================] - 69s 355ms/step - loss: 0.0193 - accuracy: 0.9071 - val_loss: 0.1077 - val_accuracy: 0.6237\n",
      "Epoch 16/20\n",
      "195/195 [==============================] - 69s 352ms/step - loss: 0.0184 - accuracy: 0.9106 - val_loss: 0.0631 - val_accuracy: 0.6827\n",
      "Epoch 17/20\n",
      "195/195 [==============================] - 70s 359ms/step - loss: 0.0175 - accuracy: 0.9141 - val_loss: 0.0810 - val_accuracy: 0.6609\n",
      "Epoch 18/20\n",
      "195/195 [==============================] - 70s 356ms/step - loss: 0.0167 - accuracy: 0.9228 - val_loss: 0.0466 - val_accuracy: 0.7692\n",
      "Epoch 19/20\n",
      "195/195 [==============================] - 70s 357ms/step - loss: 0.0152 - accuracy: 0.9316 - val_loss: 0.0737 - val_accuracy: 0.6558\n",
      "Epoch 20/20\n",
      "195/195 [==============================] - 69s 351ms/step - loss: 0.0143 - accuracy: 0.9346 - val_loss: 0.0782 - val_accuracy: 0.6462\n",
      "49/49 [==============================] - 4s 82ms/step - loss: 0.0782 - accuracy: 0.6462\n",
      "Test Accuracy: 64.61538672447205\n"
     ]
    }
   ],
   "source": [
    "\"\"\"# Train Model\r\n",
    "We initialize Adam optimizer with learning rate and decay parameters. \r\n",
    "Also, we choose the type of loss and metrics for the model and compile it for training.\r\n",
    "\"\"\"\r\n",
    "\r\n",
    "# steps_per_epoch=len(x_train) // BATCH_SIZE,\r\n",
    "\r\n",
    "# Initialize optimizer\r\n",
    "opt = Adam(learning_rate=LR, decay=LR / EPOCHS)\r\n",
    "\r\n",
    "# Compile model\r\n",
    "model.compile(loss=\"binary_crossentropy\", optimizer=opt, metrics=[\"accuracy\"])\r\n",
    "\r\n",
    "# Train model\r\n",
    "print(\"Training CNN...\")\r\n",
    "history = model.fit(augment.flow(x_train, y_train, batch_size=BATCH_SIZE),\r\n",
    "                              validation_data=(x_test, y_test),\r\n",
    "                              epochs=20, \r\n",
    "                              verbose=1)\r\n",
    "        \r\n",
    "scores = model.evaluate(x_test, y_test)\r\n",
    "print(f\"Test Accuracy: {scores[1]*100}\")\r\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEICAYAAABRSj9aAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAAA4YklEQVR4nO3de3wU1dnA8d9DQggQUCAghKBcBCwIBghQpSAqIt5ArReoF/CulXp71VK1gFikKK28ttqK1jst2loRFEVBEJHXSkS8oFCuSpD7HSGQkOf940zIZNkkm2Szu9l9vp/PfHZ35szss5PNs2fOnDkjqooxxpj4VSvaARhjjKleluiNMSbOWaI3xpg4Z4neGGPinCV6Y4yJc5bojTEmzlmiNxUiIu+IyPBwl40mEVknIgOqYbvzReQG7/mVIvJeKGUr8T7Hi8g+EUmqbKxlbFtF5MRwb9dEliX6BOAlgaKpUEQO+F5fWZFtqeq5qvpiuMvGIhEZJSILgsxPF5FDInJyqNtS1amqOjBMcZX4YVLV71U1TVUPh2P7Jv5Yok8AXhJIU9U04HvgQt+8qUXlRCQ5elHGpFeA00SkTcD8ocBXqvp1FGIypsIs0ScwEekvIrki8msR2QQ8LyKNROQtEdkqIju955m+dfzNESNEZKGITPLKrhWRcytZto2ILBCRvSIyR0SeFJFXSok7lBgfFpGPve29JyLpvuVXi8h3IrJdRB4obf+oai7wAXB1wKJrgJfKiyMg5hEistD3+mwRWS4iu0Xkz4D4lrUTkQ+8+LaJyFQROdZb9jJwPDDTOyK7T0Rae00syV6ZDBGZISI7RGSViNzo2/ZYEXlNRF7y9s0yEckubR8EfIZjvPW2evvvQRGp5S07UUQ+9D7PNhF51ZsvIvK4iGwRkT0i8lVFjoRMeFiiN82BxsAJwE2478Tz3uvjgQPAn8tYvzewAkgHHgX+JiJSibJ/Bz4FmgBjOTq5+oUS4y+Aa4FmQApwD4CIdAL+4m0/w3u/oMnZ86I/FhHpCGR58VZ0XxVtIx34N/Agbl+sBvr4iwATvPh+ArTC7RNU9WpKHpU9GuQtpgG53vqXAo+IyJm+5YO9MscCM0KJ2fMn4BigLXA67gfvWm/Zw8B7QCPc/vyTN38g0A/o4K17ObA9xPcz4aKqNiXQBKwDBnjP+wOHgNQyymcBO32v5wM3eM9HAKt8y+oBCjSvSFlckiwA6vmWvwK8EuJnChbjg77XvwTe9Z6PBqb5ltX39sGAUrZdD9gDnOa9Hg+8Wcl9tdB7fg3wia+c4BLzDaVs9yLg82B/Q+91a29fJuN+FA4DDXzLJwAveM/HAnN8yzoBB8rYtwqcCCR5+6mTb9nNwHzv+UvAFCAzYP0zgf8CPwVqRfv7n6iT1ejNVlXNK3ohIvVE5Gnv0HwPsAA4Vkrv0bGp6Imq7veeplWwbAawwzcPYH1pAYcY4ybf8/2+mDL821bVHymjhunF9E/gGu/o40pcUqvMvioSGIP6X4vIcSIyTUQ2eNt9BVfzD0XRvtzrm/cd0NL3OnDfpEr552fSgdretoJt9z7cD9anXnPQdd5n+wB3xPAksEVEpohIwxA/iwkTS/QmcPjS/wE6Ar1VtSHusBt8bcjVYCPQWETq+ea1KqN8VWLc6N+2955NylnnRVyTw9lAA2BmFeMIjEEo+Xkfwf1dunjbvSpgm2UNOfsDbl828M07HthQTkzl2Qbk45qpjtquqm5S1RtVNQNX039KvG6ZqvqEqvbAHT10AO6tYiymgizRm0ANcG3Nu0SkMTCmut9QVb8DcoCxIpIiIqcCF1ZTjP8CLhCRn4lICjCO8v8PPgJ24ZompqnqoSrG8TbQWUQu8WrSt+OasIo0APYBu0WkJUcnxs24dvKjqOp6YBEwQURSRaQrcD3uqKDS1HXdfA0YLyINROQE4O6i7YrIZb4T0TtxP0aFItJTRHqLSG3gRyAPKKxKLKbiLNGbQJOBurga3CfAuxF63yuBU3HNKL8DXgUOllJ2MpWMUVWXAbfhTqZuxCWl3HLWUVxzzQneY5XiUNVtwGXA73Gftz3wsa/IQ0B3YDfuR+HfAZuYADwoIrtE5J4gbzEM127/A/AGMEZV54QSWzl+hUvWa4CFuH34nLesJ/AfEdmHO8F7h6quARoCz+D283e4z/tYGGIxFSDeCRNjYorXPW+5qlb7EYUx8c5q9CYmeIf47USklogMAoYA06McljFxwa6ENLGiOa6JogmuKeVWVf08uiEZEx+s6cYYY+KcNd0YY0yci7mmm/T0dG3dunW0wzDGmBrls88+26aqTYMti7lE37p1a3JycqIdhjHG1Cgi8l1py6zpxhhj4pwlemOMiXOW6I0xJs7FXBu9MSby8vPzyc3NJS8vr/zCJqpSU1PJzMykdu3aIa9jid4YQ25uLg0aNKB169aUft8YE22qyvbt28nNzaVNm8A7XJbOmm6MMeTl5dGkSRNL8jFORGjSpEmFj7ws0RtjACzJ1xCV+TvFTaLfvRseeggWL452JMYYE1viJtGrwtix8NFH0Y7EGFNR27dvJysri6ysLJo3b07Lli2PvD506FCZ6+bk5HD77beX+x6nnXZaWGKdP38+F1xwQVi2FSlxczL2mGOgbl344YdoR2KMqagmTZqwdOlSAMaOHUtaWhr33FN8T5WCggKSk4Onq+zsbLKzs8t9j0WLFoUl1poobmr0IpCRYYnemHgxYsQIbrnlFnr37s19993Hp59+yqmnnkq3bt047bTTWLFiBVCyhj127Fiuu+46+vfvT9u2bXniiSeObC8tLe1I+f79+3PppZdy0kknceWVV1I0iu+sWbM46aST6NGjB7fffnu5NfcdO3Zw0UUX0bVrV37605/y5ZdfAvDhhx8eOSLp1q0be/fuZePGjfTr14+srCxOPvlkPopg80Pc1OjBEr0x4XDnneBVrsMmKwsmT674erm5uSxatIikpCT27NnDRx99RHJyMnPmzOH+++/n9ddfP2qd5cuXM2/ePPbu3UvHjh259dZbj+pz/vnnn7Ns2TIyMjLo06cPH3/8MdnZ2dx8880sWLCANm3aMGzYsHLjGzNmDN26dWP69Ol88MEHXHPNNSxdupRJkybx5JNP0qdPH/bt20dqaipTpkzhnHPO4YEHHuDw4cPs37+/4jukkuIu0S9ZEu0ojDHhctlll5GUlATA7t27GT58OCtXrkREyM/PD7rO+eefT506dahTpw7NmjVj8+bNZGZmlijTq1evI/OysrJYt24daWlptG3b9kj/9GHDhjFlypQy41u4cOGRH5szzzyT7du3s2fPHvr06cPdd9/NlVdeySWXXEJmZiY9e/bkuuuuIz8/n4suuoisrKyq7JoKibtE/9Zb7sSs9RQzpnIqU/OuLvXr1z/y/Le//S1nnHEGb7zxBuvWraN///5B16lTp86R50lJSRQUFFSqTFWMGjWK888/n1mzZtGnTx9mz55Nv379WLBgAW+//TYjRozg7rvv5pprrgnr+5YmbtrowSX6H3+EvXujHYkxJtx2795Ny5YtAXjhhRfCvv2OHTuyZs0a1q1bB8Crr75a7jp9+/Zl6tSpgGv7T09Pp2HDhqxevZouXbrw61//mp49e7J8+XK+++47jjvuOG688UZuuOEGlkSw+SGkRC8ig0RkhYisEpFRZZT7uYioiGT75v3GW2+FiJwTjqBLk5HhHq2d3pj4c9999/Gb3/yGbt26hb0GDlC3bl2eeuopBg0aRI8ePWjQoAHHHHNMmeuMHTuWzz77jK5duzJq1ChefPFFACZPnszJJ59M165dqV27Nueeey7z58/nlFNOoVu3brz66qvccccdYf8MpSn3nrEikgT8Fzgbd9PmxcAwVf0moFwD4G0gBRipqjki0gn4B9ALyADmAB1U9XBp75edna2VvfHIvHlw5pkwd657NMaE5ttvv+UnP/lJtMOIun379pGWloaqctttt9G+fXvuuuuuaId1lGB/LxH5TFWD9jMNpUbfC1ilqmtU9RAwDRgSpNzDwETAPwjDEGCaqh5U1bXAKm971aKoRr9xY3W9gzEmnj3zzDNkZWXRuXNndu/ezc033xztkMIilJOxLYH1vte5QG9/ARHpDrRS1bdF5N6AdT8JWLdlJWMtlzXdGGOq4q677orJGnxVVflkrIjUAv4I/E8VtnGTiOSISM7WrVsrHUuDBpCWZoneGGP8Qkn0G4BWvteZ3rwiDYCTgfkisg74KTDDOyFb3roAqOoUVc1W1eymTYPexDxkdtGUMcaUFEqiXwy0F5E2IpICDAVmFC1U1d2qmq6qrVW1Na6pZrCq5njlhopIHRFpA7QHPg37p/CxRG+MMSWVm+hVtQAYCcwGvgVeU9VlIjJORAaXs+4y4DXgG+Bd4LayetyEgyV6Y4wpKaQ2elWdpaodVLWdqo735o1W1RlByvb3avNFr8d763VU1XfCF3pwRYm+nF6jxpgYcsYZZzB79uwS8yZPnsytt95a6jr9+/enqCv2eeedx65du44qM3bsWCZNmlTme0+fPp1vvinuLT569GjmzJlTgeiDi6XhjOPqylhwiT4vD4L8zY0xMWrYsGFMmzatxLxp06aFNLAYuFEnjz322Eq9d2CiHzduHAMGDKjUtmJV3CX6Fi3cozXfGFNzXHrppbz99ttHbjKybt06fvjhB/r27cutt95KdnY2nTt3ZsyYMUHXb926Ndu2bQNg/PjxdOjQgZ/97GdHhjIG10e+Z8+enHLKKfz85z9n//79LFq0iBkzZnDvvfeSlZXF6tWrGTFiBP/6178AmDt3Lt26daNLly5cd911HDx48Mj7jRkzhu7du9OlSxeWL19e5ueL9nDGcTWoGZS8aKpz5+jGYkyNFIVxihs3bkyvXr145513GDJkCNOmTePyyy9HRBg/fjyNGzfm8OHDnHXWWXz55Zd07do16HY+++wzpk2bxtKlSykoKKB79+706NEDgEsuuYQbb7wRgAcffJC//e1v/OpXv2Lw4MFccMEFXHrppSW2lZeXx4gRI5g7dy4dOnTgmmuu4S9/+Qt33nknAOnp6SxZsoSnnnqKSZMm8eyzz5b6+aI9nHHc1ejtoiljaiZ/842/2ea1116je/fudOvWjWXLlpVoZgn00UcfcfHFF1OvXj0aNmzI4MHF/UW+/vpr+vbtS5cuXZg6dSrLli0rM54VK1bQpk0bOnToAMDw4cNZsGDBkeWXXHIJAD169DgyEFppFi5cyNVXXw0EH874iSeeYNeuXSQnJ9OzZ0+ef/55xo4dy1dffUWDBg3K3HYo4q5Gb003xlRRlMYpHjJkCHfddRdLlixh//799OjRg7Vr1zJp0iQWL15Mo0aNGDFiBHl5eeVvLIgRI0Ywffp0TjnlFF544QXmz59fpXiLhjquyjDHkRrOOO5q9PXru/vHWqI3pmZJS0vjjDPO4LrrrjtSm9+zZw/169fnmGOOYfPmzbzzTtkd9/r168f06dM5cOAAe/fuZebMmUeW7d27lxYtWpCfn39kaGGABg0asDfI2OYdO3Zk3bp1rFq1CoCXX36Z008/vVKfLdrDGcddjR6sL70xNdWwYcO4+OKLjzThFA3re9JJJ9GqVSv69OlT5vrdu3fniiuu4JRTTqFZs2b07NnzyLKHH36Y3r1707RpU3r37n0kuQ8dOpQbb7yRJ5544shJWIDU1FSef/55LrvsMgoKCujZsye33HJLpT5X0b1su3btSr169UoMZzxv3jxq1apF586dOffcc5k2bRqPPfYYtWvXJi0tjZdeeqlS7+lX7jDFkVaVYYqLDBgA+/dDAt/03ZgKsWGKa5bqGKa4xrEavTHGFIvLRN+iheteGWMHK8YYExVxmegzMuDQIdixI9qRGFNzxFozrgmuMn+nuE30YM03xoQqNTWV7du3W7KPcarK9u3bSU1NrdB6cdvrBlyi79IlurEYUxNkZmaSm5tLVW78YyIjNTWVzMzMCq0T94neGFO+2rVr06ZNm2iHYapJXDbd2NWxxhhTLC4TfWoqNG5sid4YYyDERC8ig0RkhYisEpFRQZbfIiJfichSEVkoIp28+a1F5IA3f6mI/DXcH6A01pfeGGOcctvoRSQJeBI4G8gFFovIDFX1DyH3d1X9q1d+MPBHYJC3bLWqZoU16hBYojfGGCeUGn0vYJWqrlHVQ8A0YIi/gKru8b2sD0S9j1aLFpbojTEGQkv0LYH1vte53rwSROQ2EVkNPArc7lvURkQ+F5EPRaRvsDcQkZtEJEdEcsLVvSsjAzZtgsLCsGzOGGNqrLCdjFXVJ1W1HfBr4EFv9kbgeFXtBtwN/F1EGgZZd4qqZqtqdtOmTcMST0YGFBSAd3cxY4xJWKEk+g1AK9/rTG9eaaYBFwGo6kFV3e49/wxYDXSoVKQVZH3pjTHGCSXRLwbai0gbEUkBhgIz/AVEpL3v5fnASm9+U+9kLiLSFmgPrAlH4OWxRG+MMU65vW5UtUBERgKzgSTgOVVdJiLjgBxVnQGMFJEBQD6wExjurd4PGCci+UAhcIuqRmSoMUv0xhjjhDQEgqrOAmYFzBvte35HKeu9DrxelQArq3lz92iJ3hiT6OLyyliAlBRo2tQSvTHGxG2iB+tLb4wxEOeJPiPD3WnKGGMSWdwneqvRG2MSXdwn+k2b4PDhaEdijDHRE/eJvrAQtmyJdiTGGBM9cZ/owZpvjDGJzRK9McbEOUv0xhgT5+I60R93HIhYojfGJLa4TvTJydCsmSV6Y0xii+tED3bRlDHGJESitxq9MSaRWaI3xpg4lxCJfssWyM+PdiTGGBMdISV6ERkkIitEZJWIjAqy/BYR+UpElorIQhHp5Fv2G2+9FSJyTjiDD0VGBqjC5s2RfmdjjIkN5SZ671aATwLnAp2AYf5E7vm7qnZR1SzgUeCP3rqdcLce7AwMAp4qurVgpFhfemNMogulRt8LWKWqa1T1EO7m30P8BVR1j+9lfUC950OAad5NwtcCq7ztRYwlemNMogvlVoItgfW+17lA78BCInIbcDeQApzpW/eTgHVbBln3JuAmgOOPPz6UuEPWooV7tERvjElUYTsZq6pPqmo74NfAgxVcd4qqZqtqdtOmTcMVEuAumKpVyxK9MSZxhZLoNwCtfK8zvXmlmQZcVMl1wy4pyd0o3C6aMsYkqlAS/WKgvYi0EZEU3MnVGf4CItLe9/J8YKX3fAYwVETqiEgboD3wadXDrhjrS2+MSWTlttGraoGIjARmA0nAc6q6TETGATmqOgMYKSIDgHxgJzDcW3eZiLwGfAMUALepasTv95SRAevWRfpdjTEmNoRyMhZVnQXMCpg32vf8jjLWHQ+Mr2yA4ZCRAYsWRTMCY4yJnri/MhZcot+2DQ4ejHYkxhgTeQmT6MHdKNwYYxJNQiV6OyFrjElECZHo7aIpY0wiS4hEX1Sjt770xphElBCJPj3d3VbQavTGmESUEIm+Vi3XfGOJ3hiTiBIi0YNdHWuMSVyW6I0xJs5ZojfGmDiXUIl+5044cCDakRhjTGQlTKIv6ktvXSyNMYkmYRK9XR1rjElUCZforUZvjEk0CZforUZvjEk0ISV6ERkkIitEZJWIjAqy/G4R+UZEvhSRuSJygm/ZYRFZ6k0zAteNlMaNISXFEr0xJvGUe+MREUkCngTOBnKBxSIyQ1W/8RX7HMhW1f0icivwKHCFt+yAqmaFN+yKE7EulsaYxBRKjb4XsEpV16jqIdzNv4f4C6jqPFXd7738BHcT8Jhjid4Yk4hCSfQtgfW+17nevNJcD7zje50qIjki8omIXBRsBRG5ySuTs3Xr1hBCqhxL9MaYRBTWk7EichWQDTzmm32CqmYDvwAmi0i7wPVUdYqqZqtqdtOmTcMZUgk2sJkxJhGFkug3AK18rzO9eSWIyADgAWCwqh65O6uqbvAe1wDzgW5ViLdKMjJgzx7Yty9aERhjTOSFkugXA+1FpI2IpABDgRK9Z0SkG/A0Lslv8c1vJCJ1vOfpQB/AfxI3oqwvvTEmEZWb6FW1ABgJzAa+BV5T1WUiMk5EBnvFHgPSgH8GdKP8CZAjIl8A84DfB/TWiShL9MaYRFRu90oAVZ0FzAqYN9r3fEAp6y0CulQlwHCyi6aMMYkoYa6MBUv0xpjElFCJ/phjoG5dS/TGmMSSUInero41xiSihEr0YIneGJN4Ei7R20VTxphEk3CJ3mr0xphEk5CJ/scfYe/eaEdijDGRkZCJHqxWb4xJHJbojTEmzlmiN8aYOGeJ3hhj4lzCJfoGDSAtzRK9MSZxJFyiB+tLb4xJLAmZ6K0vvTEmkViiN8aYOJewiX7jRlCNdiTGGFP9Qkr0IjJIRFaIyCoRGRVk+d0i8o2IfCkic0XkBN+y4SKy0puGhzP4ysrIgAMHYPfuaEdijDHVr9xELyJJwJPAuUAnYJiIdAoo9jmQrapdgX8Bj3rrNgbGAL2BXsAYEWkUvvArx7pYGmMSSSg1+l7AKlVdo6qHgGnAEH8BVZ2nqvu9l58Amd7zc4D3VXWHqu4E3gcGhSf0yrNEb4xJJKEk+pbAet/rXG9eaa4H3qnIuiJyk4jkiEjO1q1bQwipaizRG2MSSVhPxorIVUA28FhF1lPVKaqararZTZs2DWdIQbVo4R4t0RtjEkEoiX4D0Mr3OtObV4KIDAAeAAar6sGKrBtp9etDw4aW6I0xiSGURL8YaC8ibUQkBRgKzPAXEJFuwNO4JL/Ft2g2MFBEGnknYQd686LO+tIbYxJFcnkFVLVAREbiEnQS8JyqLhORcUCOqs7ANdWkAf8UEYDvVXWwqu4QkYdxPxYA41R1R7V8kgqyRG+MSRTlJnoAVZ0FzAqYN9r3fEAZ6z4HPFfZAKtLRgYsXBjtKIwxpvol5JWxUFyjt6tjjTHxLqET/aFDsCMmGpKMMab6JHSiB2unN8bEP0v0luiNMXEuYRO9XTRljEkUlugt0Rtj4lzCJvq6daFRI0v0xpj4l7CJHopvQGKMMfEs4RO91eiNMfHOEr0lemNMnEv4RL9xIxQWRjsSY4ypPgmf6AsKYNu2aEdijDHVJ+ETPVjzjTEmviV0ore+9MaYRJDQid5q9MaYRBBSoheRQSKyQkRWicioIMv7icgSESkQkUsDlh0WkaXeNCNw3Whq3tw9WqI3xsSzcm88IiJJwJPA2UAusFhEZqjqN75i3wMjgHuCbOKAqmZVPdTwq1MH0tPtoiljTHwL5Q5TvYBVqroGQESmAUOAI4leVdd5y2pcR0XrS2+MiXehNN20BNb7Xud680KVKiI5IvKJiFwUrICI3OSVydm6dWsFNl11luiNMfEuEidjT1DVbOAXwGQRaRdYQFWnqGq2qmY3bdo0AiEVs0RvjIl3oST6DUAr3+tMb15IVHWD97gGmA90q0B81S4jAzZtgsOHox2JMcZUj1AS/WKgvYi0EZEUYCgQUu8ZEWkkInW85+lAH3xt+7GgRQs3BMKWLdGOxBhjqke5iV5VC4CRwGzgW+A1VV0mIuNEZDCAiPQUkVzgMuBpEVnmrf4TIEdEvgDmAb8P6K0TddaX3hgT70LpdYOqzgJmBcwb7Xu+GNekE7jeIqBLFWOsVv5E36NHdGMxJqpUQSTaUZhqkNBXxoLV6I05YuBA6N/fRvmLQwmf6I87zlVi7KIpk9DWr4c5c+DDD+HUU2HlymhHZMIovhJ9JbrO1K4NzZpZjd4kuLfeco/PPQe7drlk//HHUQ3JhE/8JPp9+6BNG7j9dlixokKrWl96k/DeegvatYMRI+CTT6BJEzjrLHj11WhHZsIgfhL93r1w+unw9NNw0kmuvXHGjJBq+ZboTUL78UeYOxcuuMC1Y7ZrB4sWQc+eMHQoTJzoTtSaGit+En2LFvDyy/D99/C738G338KQIe5LO3FimSeYLNGbhDZnDhw8CBdeWDyvSRN4/32X6EeNgltucbdjMzVS/CT6IscdBw88AGvXwuuvQ9u27ouamekOS3NyjlqlRQt3wVR+fuTDNSbq3noLGjaEvn1Lzk9NhalT4f77YcoU90Owd290YjRVEn+JvkhyMlxyCXzwAXz9NVx/vUv8PXtC796u9p+XB7gavSps3hzlmI2JtMJCl+jPOQdSUo5eXqsWjB8Pzzzjavh9+0JubuTjNFUSv4ner3NnePJJ2LAB/vQn2L0brrkGjj8e7r+fdrW/B6z5xiSgJUvcYE8XXFB2uRtugFmzYM0aV1FaujQi4ZnwSIxEX6RhQxg50rXfv/8+9OkDEydy1o1teJ1LyH93rp10Moll5kxXaz/vvPLLDhwICxe68n37wrvvVn98JiwSK9EXEYEBA+CNN2DNGn785X305SP6jBkAN93kDmeNSQRvveX6zKenh1a+a1fX/fLEE91RwJQp1RufCYvETPR+J5xA3ckTOEHWs+C0X8Ozz8Ivf2nJ3sS/DRtc0015zTaBWraEBQtcu/7NN7vODvb/EtMs0QNJSdCoRSovdpzgvrRPP+2aeKwZx8Szt992j/5ulaFq0ADefNN1u5w4EYYNO9K5wcSekEavTAQZGfDDRoG/PeIusnrsMddz53//10b0M/Fp5kxo3Ro6darc+snJ8NRTrgvzffe5I4Tp00NvBjIRY4ne06IFfPcdLqlPnOguDnn8cVfd/+MfLdmb+LJ/v7tQ6sYbq/bdFoF773U/GFdfDeee68bICdZV00RNSE03IjJIRFaIyCoRGRVkeT8RWSIiBSJyacCy4SKy0puGhyvwcCtxdawI/OEPbtycyZPdF9macUw8+eAD19RS0fb50lx2GfzjH+6CxFFHpQgTZeXW6EUkCXgSOBvIBRaLyIyAO0V9D4wA7glYtzEwBsgGFPjMW3dneMIPn4wMN0rCwYNQpw4u2U+e7Jpx/vAHd5g6YYLV7E18eOstSEtz40OFy8UXu3Nbjz8OZ5xRubZ/Uy1CqdH3Alap6hpVPQRMA4b4C6jqOlX9Egg89X4O8L6q7vCS+/vAoDDEHXZFNyDZtMk3U8RdYFV0wunBB61mb2o+VZfoBw70ajVh9NhjkJXlhhtZvz682zaVFkqibwn4/2K53rxQhLSuiNwkIjkikrN169YQNx1eRYn+qBuQiLiram+8ER55BMaOjXRoxoTX0qXuxGl11LhTU93QxocOuZ44NhBaTIiJ7pWqOkVVs1U1u2nTplGJocxbCtaqBX/9K1x3HYwb5yZjaqqZM10FJpSrYSujQwf3//Lxx1YxihGhJPoNQCvf60xvXiiqsm5EZXq3Nn/hhVIG6KtVyw3sNHw4jBnjBnoypiaaOdONV9OsWfW9x5VXuorRI4+43j0mqkJJ9IuB9iLSRkRSgKHAjBC3PxsYKCKNRKQRMNCbF3PS0+HRR901JNnZ8OWXQQrVqgV/+xtcdZVrr584MeJxGlMlGze6njHh6m1TlieecDcBuuoqGxo2yspN9KpaAIzEJehvgddUdZmIjBORwQAi0lNEcoHLgKdFZJm37g7gYdyPxWJgnDcvJt17L8yb52r0vXu7CvxR516Tkly1f9gw141s0qRohGpM5VTlatiKql8fXnvNjRZ71VU2TEI0qWpMTT169NBo27xZdeBAVVD9xS9U9+wJUig/X/Xyy12hP/4x4jEaUylDhqi2aqVaWBi595wyxf2fjB8fufdMQECOlpJXY+JkbKxp1gzeecfdkXDatFKacpKT4ZVX4Oc/h7vvdoeppmY7eDC+u8/m5bnhuS+8MLLXg9xwA1xxBYwe7YY5NhFnib4UtWq5OxJ+8EFxU86zzwbkgdq13dWAF18Md9zhumGamum779yYLRddFL9dAufNc0MfRKJ93k/EDWfcurVr8ty+PbLvbyzRl+f001234379XFf6q6+Gfft8BWrXdtX+wYPdVYHDhrk78cRrsgik6vpkz5vnfglnzqx5J9727HHJb+dOmDEDbrstPmv2M2e6dvMzzoj8ezds6P5PNm+Ga6+tnv27eTPsiNlTgFFlg5qFoKgpZ8IEd/SZk+POMXXt6hVISXEzRo2CF190X+imTWHoUHcSqmfP6j1UVnX3xZ0/371Ps2buJunNmrmpUSN3iFKV7W/fDv/9L6xc6R6Lnq9c6WqJgY4/3n3uXr3cY48e7p891hw+DL/4hbvr2LvvukO4CRPghBPcTbHjRdHVsGef7S5qiobsbHfl7J13ulFh77wzPNvdudN14/zTn9z3fsEC9/czR4jGWM0lOztbc3Jyoh1GqT780FXad+5036vrrw/I4YcOuV+FV15xNaiDB6F9e5fwr7wS2rULTyCbNrn+ye+959pdS4zdECA52f3w+JN/0XP/vKZNYevWkgm96HHXruLtJSW5Zo727d3FMUWPbdu6K84+/dRNixe7e4yC20knneQSf1Hy79o1/JfgV9Tdd7uxWf7yFzfUhaq7n/Arr8BLL7lDuHjwxRduaIJnn3Vf2mhRdc1j77wDixa55F9ZeXnun/CRR1zPnqFD3XabNHHJvugqyFhTWOiOglevdv8fq1cXTyecAP/6V6U2KyKfqWrQHWqJvhK2bHF5+/33Xe7+61/d+FBH2bULXn8dpk51tW1Vd9u2q66Cyy+v2LjdBw7ARx+5N33vveKzw02auFra2WfDWWdB3bruEHbLFjcVPQ+ct3lz2TeKEIFWrUom8qLH1q1dk1Uotm1zh0D+5L9li1uWkgKnnFKc+Hv1cj8GkTpROGWKu0PS7be7GmaRQ4fccLsLFrha/llnRSae6jR+vLv2Y+NGaN48urFs3+5+dFJS3B2ujjmmYusfPuz+px580I2nc9558PvfQ5cu8J//uNuEtmrlamVRutKegwdh7dqSSbxoWrvWLS+SlOQSfLt2Lj889FCl3rKsRB/17pSBUyx0rwzF4cOqv/udaq1aqh07qn7xRTkrfP+96sSJqief7LqaJSerXnih6quvqu7fH/wNli5VffRR1bPPVq1Tx62XkqJ6xhmqEyao5uS4cpVRWOj6ja5apbpokeobb7hucP/+t+pXXwWPKRwKC1XXrVP95z9V771XtX9/1bQ099lA9YILVPfurZ739pszx/0Nzj3XdZUNtGuX+1s1bBjCH7cG6N1btWfPaEdR7KOPVJOSVK+4IvSunoWFqu+8o9q1q/uuZGerfvDB0eXmz1etW1c1K0t1x47wxl1WbP/7v6qnn66amakqUvydBtX69V3cF1+ses89qn/5i+p776muXq166FBYQqCM7pVRT+yBU01J9EXmz1dt0UI1NVX1mWdC/M5+8YVLci1buj9Bgwaq116r+u67qi++qHrllarNmhV/STp3Vr3rLtVZs1T37av2zxRxBQWqX3/t+lnXqqXavbvqxo3V937Ll6see6zbr7t3l17u++/d36hlS9X166svnuq2aZNLPOPGRTuSksaPd9/vKVPKL5uTo3rmma5827aq06aVXcl5911XKerdu5QLYcLoxx9Vhw51sXXrpnr11apjx6q+/LKrRG3eHJHrFizRV7PNm12lG1wlcNw41W+/DWHFggLVuXNdkm/YsDixN2vmkv0LL6jm5lZ7/DFl5kzVevVUTzhBddmy8G9/2zbVE09UbdpUdc2a8st/8YX725x8sqvl10TPPee+V0uWRDuSkg4fdv84qamqX34ZvMzq1cVJND1d9YknVA8eDG3706e7o4Z+/Vwyrg7r16v26OF+SH//+8heiBbAEn0EHD7sKiZ9+xYftXXpovrww64CWa79+1Xffts111S2OSZeLF6setxxrtY9f374tnvwoGsqSklR/fjj0NcrauY588zQk0wsueQSd1QSxSRUqk2b3N/6Jz8pebS6davqHXeo1q7tmmEefLDso6/S/OMf7h9y4EDVvLywha2qqv/3f6rNm7umxxkzwrvtSrBEH2EbNriKx89+Vpz0u3Z1bforVkQ7uhpi7Vr3z1+7turUqVXfXmGh6vXXuz/GK69UfP2XXnLrXnVVbCbM0uTluUR0883RjqR077/v/lGuvdbVvMePd0dRtWqp3nij+4eqiqIjmiFDwtYeri+84CoMbdu6ZscYYIk+inJz3TmaPn30SMvMKae47/J//xvt6GLcjh3u5BaoPvJI1RLspEluOw8+WPlt/O53bhv331/5bUTau++6mN96K9qRlO2BB1ycTZoUJ+Vvvgnf9v/8Z7fdoUNdk2llFRSo/s//uG2dcYZrCowRluhjxPr1qpMnq552WnHSz8pyOWzlymhHF6Py8tzIcuBqd8F6yJTnzTddjfHSS6vWLFZY6GIA1aefrvx2ImnkSNf0UV29qMIlP1/1/PNd2+dHH1XPe0yc6P52115bue/Bzp2qgwa5bYwcGb6jgzCxRB+D1q9Xffzxo5P+b3+rumBBzH2HoquwsLjGN2hQxXpRLF3qurZlZ4fnhFx+vup557lmhZkzq7696lRY6E5qX3hhtCOJHaNHu+/RbbdV7AhxxQrXjzo5OWZ/5C3Rx7jvv3cjHZ92mssf4JpVL7jANft8803NahauNlOmuF4UWVmhtdtu3OiG5G3ZUvWHH8IXx969rqdFvXqqn34avu2G21df1ayjj0goLHT92EH1vvtC+8eaPdt1DEhPV/3ww+qPsZKqnOiBQcAKYBUwKsjyOsCr3vL/AK29+a2BA8BSb/pree+ViIneb+dOd83Srbe6XoBFtf3MTNURI1T//nfXnTNhzZrlfgVbtXKJrDT796v26uWScXV0K9y4UbV1a9cVdvXq8G8/HB55xH15Eq2LbnkKC1V/+Uu3bx56qOxyjz/ual9du7oOAjGsSokeSAJWA22BFOALoFNAmV8WJXHcrQZf1eJE/3V57+GfEj3RB1q71lVkL71UtVGjks08997rLq6L9ebXsFuyxF2l1rChuw4hUGGhu+JSxF3xW12WL1dt3Fi1Q4eYOil3xGmnuYvPzNEOH3Y1J3An6gPl5bm2fHBXs0biau0qqmqiPxWY7Xv9G+A3AWVmA6d6z5OBbYBYog+vggLXUjB+vOsOXru2+wumpqoOGODONS1cGJ8Xzx7lu+/cla21a7uuj35jxrgdM3Fi9cexcKEbnuK002LrF3fLFvdDN2ZMtCOJXQUFxXeJe/LJ4vmbNhWfPBs9usZc11LVRH8p8Kzv9dXAnwPKfA1k+l6vBtK9RP8j8DnwIdC3vPezRB+6fftcS8add7qcV1Tbr1XLXcg5fLjqn/7kruuIpRwUNjt3Fl8WP26cq8lPnapHelZE6sTGP//pkurPf161rnvh9OKLbj8sXhztSGLboUPuZDWoPv+8O1rMzHQ9lV57LdrRVUhZib7c0StF5FJgkKre4L2+GuitqiN9Zb72yuR6r1cDvYG9QJqqbheRHsB0oLOq7gl4j5uAmwCOP/74Ht99912ZMZngNm50g0N+9pkbMDInp3igyKQk6NzZjQrbo4d77No1ekOTh82hQ+6OMC+9VDz8be/ebpTPlJTIxTF5Mtx1lxvZs1UrN0JkixZHT82buxFHq3J/gFBcdhl8/DHk5lb/e9V0eXnuxkFz57phs9PT4c03oVu3aEdWIVUaplhETgXGquo53uvfAKjqBF+Z2V6Z/xORZGAT0FQDNi4i84F7VLXUcYhrwjDFNYWqG/a6KOkX/QBs2+aWJyfDySeXTP6dOkG9etGNu8JUYexYGDfOjYn/n/9UbAjocHn6aXenrU2b3K/uxo3uPpSBkpNL/hD4nw8YUPV7Fhw65D7/0KFuKGZTvv373S1B8/Pd7UGPOy7aEVVYVRN9MvBf4CxgA7AY+IWqLvOVuQ3ooqq3iMhQ4BJVvVxEmgI7VPWwiLQFPvLKlXq/L0v01UsVvv++ZK3/s89K3oEtM9MNOx84VWQY+qiYN8+NmZ+ZGe1Iiv34Y3HS9/8A+KdNm9wNX8D9CPzyl/Db31b+x2rOHHd/gjffdDVVkxCqfOMRETkPmIzrgfOcqo4XkXG4NqEZIpIKvAx0A3YAQ1V1jYj8HBgH5AOFwBhVnVnWe1mijzxVWLfOJfzly4vvFLhiRckbSyUnuwpzsB+BjIzI3S8kLuXnu1/gxx6DZ56BBg3c3el/9auKt6/deac7uti+vQYenpnKsjtMmUpRLb5VbOC0cmXJG1TVq1cy8XfsWDzF4q1iY9qyZfDrX8Pbb7s7D02YAFdcEVpbuyqceKK7U9fbb1d/rCZmWKI3YVd020t/8l+xwj2uXeuWF2ne/Ojk36EDtGkT401B0TZ3LtxzDyxd6m61OGkS9OtX9jrffOPOuhfdA9ckjLISfXKkgzHxoVYt17mkVaujb6l68KC7NWZR8i+a3nij+EQwuKagdu2KE3/HjtCyJTRu7KYmTdztRJOSIvvZYsZZZ7n2tFdegfvvh9NPdz2LJk50OyyYt95yj+efH7EwTeyzGr2JqB07ihO//4dg5UrXWSSQCBx7rEv6/h+A0p4fe6ybjjnG9ZSLG/v3uy6cEya4NrNbboHRo4+++XXfvrBvH3z+eVTCNNFjTTcm5h0+7M5Fbt7sfgy2b3ePgc/9r/0nioNJTXUJvyjxl/VY9Dw93fWsa9QoRk8ub94MDz3kuk3Wr+9q+rffDnXruh3TrJmb9/DD0Y7URJglehOXDh+GnTtL/gDs2uWm3bvLfty1q+TJ5EDJyS5nHnfc0VPg/PT0KDQvffutO2E7cyYcfzw88og7MXLNNe46gl69IhyQiTZL9MYEcfCgS/z+5L9tm6s0B05btrjHYM1LtWq5ZN+smZvS010zUnp68RT4ul69MB0xzJvnTtguWeKuBG7UCH74wa6GTUB2MtaYIOrUKU7OoVB1Pwql/Qhs3uyue/riC/eDsWOHW6e09w72Q9CwoWuRKZrq1SvndZ8zSFm8GP7+dxgzxg19YEneBLAavTHV5PDh4qOEbdtc01LR88DXRc/37HHXTlVEcnJx4q9bt/wpNbX0ZQ0auIMC/2RdYGsGq9EbEwVJSa6m3qSJ6zoaqvx8N3LC/v3u0T+VN2//fjhwwJ1/OHDAncP44Qf3PHAKVf36Ryf/0qa0NHe0kpLiHoNNKSkxeqI7jlmiNybG1K5d3BOouqi6cxSByX/vXvfjUNa0Zk3x8x9/rNz7165d+g+Bf0pNDX1+0byUlMpP8XrNhiV6YxKQiEuMqamuJl5Z+fmuecqf+A8eLH06dKjs5f5p3z7XpBU4Py+veFvhlpzsjkr850GCTaWVqVfv6B+iouf+x0gf1ViiN8ZUWu3a7pqtwOu2IkG15A9HXp57XZGp6Aej6Hle3tFNYz/+6E6sr19fcl5Fmr+CCfZD0KOHGyU53CzRG2NqJJHiZBkNhYUlz4/s2+eSv/+ow/+8tEf/8zZtqidWS/TGGFMJtWq5Jpy0tGhHUj7rcGuMMXHOEr0xxsS5kBK9iAwSkRUiskpERgVZXkdEXvWW/0dEWvuW/cabv0JEzglj7MYYY0JQbqIXkSTgSeBcoBMwTEQ6BRS7HtipqicCjwMTvXU7AUOBzsAg4Clve8YYYyIklBp9L2CVqq5R1UPANGBIQJkhwIve838BZ4mIePOnqepBVV0LrPK2Z4wxJkJCSfQtgfW+17nevKBlVLUA2A00CXFdROQmEckRkZytW7eGHr0xxphyxcTJWFWdoqrZqprdNBpXXhhjTBwLJdFvAFr5Xmd684KWEZFk4Bhge4jrGmOMqUblDlPsJe7/AmfhkvRi4BequsxX5jagi6reIiJDgUtU9XIR6Qz8HdcunwHMBdqr6uEy3m8r8F3VPla1Sge2lVsqeiy+qrH4qsbiq5qqxHeCqgZtEin3ylhVLRCRkcBsIAl4TlWXicg4IEdVZwB/A14WkVXADlxPG7xyrwHfAAXAbWUleW+dmG67EZGc0sZ8jgUWX9VYfFVj8VVNdcUX0hAIqjoLmBUwb7TveR5wWSnrjgfGVyFGY4wxVRATJ2ONMcZUH0v0FTcl2gGUw+KrGouvaiy+qqmW+GLunrHGGGPCy2r0xhgT5yzRG2NMnLNEH0BEWonIPBH5RkSWicgdQcr0F5HdIrLUm0YH21Y1x7lORL7y3j8nyHIRkSe8kUO/FJHuEYyto2/fLBWRPSJyZ0CZiO5DEXlORLaIyNe+eY1F5H0RWek9Br17qogM98qsFJHhEYzvMRFZ7v393hCRY0tZt8zvQjXGN1ZENvj+hueVsm6Zo99WY3yv+mJbJyJLS1k3EvsvaF6J2HdQVW3yTUALoLv3vAHuYrFOAWX6A29FOc51QHoZy88D3gEE+CnwnyjFmQRswl3MEbV9CPQDugNf++Y9Cozyno8CJgZZrzGwxnts5D1vFKH4BgLJ3vOJweIL5btQjfGNBe4J4e+/GmgLpABfBP4/VVd8Acv/AIyO4v4Lmlci9R20Gn0AVd2oqku853uBbwkyEFsNMAR4SZ1PgGNFpEUU4jgLWK2qUb3aWVUX4C7m8/OPuvoicFGQVc8B3lfVHaq6E3gfN+R2tcenqu+pGyQQ4BPcECJRUcr+C0Uoo99WWVnxeSPpXg5Uw223Q1NGXonId9ASfRnE3UClG/CfIItPFZEvROQdb6iHSFPgPRH5TERuCrI8pJFDI2Aopf+DRXsfHqeqG73nm4DjgpSJlf14He4ILZjyvgvVaaTXtPRcKc0OsbD/+gKbVXVlKcsjuv8C8kpEvoOW6EshImnA68CdqronYPESXFPEKcCfgOkRDg/gZ6raHXdDmNtEpF8UYiiTiKQAg4F/BlkcC/vwCHXHyDHZ11hEHsANITK1lCLR+i78BWgHZAEbcc0jsWgYZdfmI7b/ysor1fkdtEQfhIjUxv0xpqrqvwOXq+oeVd3nPZ8F1BaR9EjGqKobvMctwBscfUOXWBg59FxgiapuDlwQC/sQ2FzUnOU9bglSJqr7UURGABcAV3qJ4CghfBeqhapuVtXDqloIPFPK+0Z7/yUDlwCvllYmUvuvlLwSke+gJfoAXnve34BvVfWPpZRp7pVDRHrh9uP2CMZYX0QaFD3HnbT7OqDYDOAacX4K7PYdIkZKqTWpaO9DzwygqAfDcODNIGVmAwNFpJHXNDHQm1ftRGQQcB8wWFX3l1ImlO9CdcXnP+dzcSnvuxhoLyJtvCO8obj9HikDgOWqmhtsYaT2Xxl5JTLfweo801wTJ+BnuMOnL4Gl3nQecAtwi1dmJLAM14PgE+C0CMfY1nvvL7w4HvDm+2MU3L1+VwNfAdkRjrE+LnEf45sXtX2I+8HZCOTj2jivx90FbS6wEpgDNPbKZgPP+ta9DncbzFXAtRGMbxWubbboe/hXr2wGMKus70KE4nvZ+259iUtYLQLj816fh+tlsjqS8XnzXyj6zvnKRmP/lZZXIvIdtCEQjDEmzlnTjTHGxDlL9MYYE+cs0RtjTJyzRG+MMXHOEr0xxsQ5S/TGGBPnLNEbY0yc+3+AmxUp8DYXuQAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] Calculating model accuracy\n",
      "49/49 [==============================] - 4s 88ms/step - loss: 0.0782 - accuracy: 0.6462\n",
      "Test Accuracy: 64.61538672447205\n"
     ]
    }
   ],
   "source": [
    "\"\"\"# Evaluate Model\r\n",
    "Comparing the accuracy and loss by plotting the graph for training and validation.\r\n",
    "\"\"\"\r\n",
    "\r\n",
    "acc = history.history['accuracy']\r\n",
    "val_acc = history.history['val_accuracy']\r\n",
    "loss = history.history['loss']\r\n",
    "val_loss = history.history['val_loss']\r\n",
    "epochs = range(1, len(acc) + 1)\r\n",
    "\r\n",
    "# Train and validation accuracy\r\n",
    "plt.plot(epochs, acc, 'b', label='Training accurarcy')\r\n",
    "plt.plot(epochs, val_acc, 'r', label='Validation accurarcy')\r\n",
    "plt.title('Training and Validation accurarcy')\r\n",
    "plt.legend()\r\n",
    "\r\n",
    "plt.figure()\r\n",
    "\r\n",
    "# Train and validation loss\r\n",
    "plt.plot(epochs, loss, 'b', label='Training loss')\r\n",
    "plt.plot(epochs, val_loss, 'r', label='Validation loss')\r\n",
    "plt.title('Training and Validation loss')\r\n",
    "plt.legend()\r\n",
    "plt.show()\r\n",
    "\r\n",
    "\"\"\"Evaluating model accuracy by using the `evaluate` method\"\"\"\r\n",
    "\r\n",
    "print(\"[INFO] Calculating model accuracy\")\r\n",
    "scores = model.evaluate(x_test, y_test)\r\n",
    "print(f\"Test Accuracy: {scores[1]*100}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Saving the Model\r\n",
    "\r\n",
    "# model.save('PDD_completemodel')\r\n",
    "# model.save('PDD_completemodel.h5')\r\n",
    "# intermediate_layer_model.save('PDD_IntermediateModel')\r\n",
    "# intermediate_layer_model.save('PDD_IntermediateModel.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "id": "Xr2hsdio8hkz"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1560, 1024)\n"
     ]
    }
   ],
   "source": [
    "x_train_predict = intermediate_layer_model.predict(x_train)\r\n",
    "# print(x_train_predict.shape)\r\n",
    "\r\n",
    "x_test_predict = intermediate_layer_model.predict(x_test)\r\n",
    "print(x_test_predict.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Support Vector Machine\r\n",
    "\r\n",
    "Integration of CNN with Support Vector Machine."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SVM Fit Complete\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import SVC\r\n",
    "\r\n",
    "svm = SVC(kernel='rbf')\r\n",
    "\r\n",
    "svm.fit(x_train_predict,np.argmax(y_train,axis=1))\r\n",
    "\r\n",
    "print('SVM Fit Complete')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.986698717948718"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "svm.score(x_train_predict,np.argmax(y_train,axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9634615384615385"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "svm.score(x_test_predict,np.argmax(y_test,axis=1))\r\n",
    "\r\n",
    "#Save the SVM model in pickle file\r\n",
    "# pickle.dump(svm,open('svms.pkl', 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    0\n",
       "0  20\n",
       "1  19\n",
       "2  10\n",
       "3  14\n",
       "4   1"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Pred_labels = svm.predict(x_test_predict)\r\n",
    "Pred_labels = pd.DataFrame(Pred_labels,index =None)\r\n",
    "Pred_labels.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# XGBoost Integration\r\n",
    "\r\n",
    "Integrating the CNN model with Extreme Gradient Boost algorithm."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[01:15:35] WARNING: C:/Users/Administrator/workspace/xgboost-win64_release_1.4.0/src/learner.cc:1095: Starting in XGBoost 1.3.0, the default evaluation metric used with the objective 'multi:softprob' was changed from 'merror' to 'mlogloss'. Explicitly set eval_metric if you'd like to restore the old behavior.\n",
      "XGBoost Fit Complete\n"
     ]
    }
   ],
   "source": [
    "import xgboost as xgb\r\n",
    "\r\n",
    "xb = xgb.XGBClassifier(use_label_encoder=False)\r\n",
    "\r\n",
    "xb.fit(x_train_predict,np.argmax(y_train,axis=1))\r\n",
    "\r\n",
    "print('XGBoost Fit Complete')\r\n",
    "\r\n",
    "#Save XGBoost Model\r\n",
    "# pickle.dump(xb,open('xgb_model.pkl','wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xb.score(x_train_predict,np.argmax(y_train,axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9756410256410256"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xb.score(x_test_predict,np.argmax(y_test,axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from tensorflow.keras.models import load_model\r\n",
    "\r\n",
    "# ai = load_model('D:\\PDD_Models\\HDF5\\PDD_Intermediatemodel.h5')\r\n",
    "# xgb_model = xgb.XGBClassifier()\r\n",
    "# xgb_model.load_model(r'D:\\PDD_Models\\xgb_model.json')\r\n",
    "\r\n",
    "# x2 = ai.predict(x_test)\r\n",
    "# xgb_model.score(x_test_predict,np.argmax(y_test,axis=1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Further Scope of Improvement\r\n",
    "\r\n",
    "- Image processing techniques can be implemented before approaching the CNN's input layer\r\n",
    "- Developing a Real-Time PDD [Plant Disease Detection] system using Opencv\r\n",
    "- Improving the Validation Loss \r\n",
    "- Wider scope of data augmentation & manipulation for better performance\r\n",
    "- Integrating CNN with other architectures\r\n",
    "- Hyperparameter tunning\r\n",
    "- Additional Datasets\r\n",
    "- Front End development [Django, Flask, HTML+CSS+JS, Kotlin, Java, Swift, Flutter]\r\n",
    "- Cross Platform application development \r\n",
    "- Deploying the AI model onto Companion computers for real-time application "
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "authorship_tag": "ABX9TyMYnTMjyEbKj3D0r2HxMd9C",
   "name": "Plant_Disease_Detection.ipynb",
   "provenance": []
  },
  "interpreter": {
   "hash": "64c1db036c245e91d1c2d164e41aef3f7526caa4d16da4bd1475087c34dc1c7b"
  },
  "kernelspec": {
   "display_name": "Python 3.8.10 64-bit ('DL': conda)",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": ""
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}