FF_VGG16_Model_Brain_Tumor_Classification_+_Model_Creation_and_Fine_Tunning_.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_AhrXk3clJUN"
   },
   "source": [
    "## The VGG16 model development and fine tuning\n",
    "\n",
    "in this notebook we will load and add our layers to the base model of the pre-trained VGG16 model, and fine tune the following components of the model:\n",
    "1. **Optimizers** - We will use only 5 optimizers to see which one works best, namely; 1. `SGD`,2. `RMSprop`,3. `Adam`,4. `Adagrad`,5. `Adadelta`. the best optimizer will be used in the New model\n",
    "2. **Number of epochs** - We will also use only 4 epochs to choose the best performer namely; `1, 2, 5, 10`.\n",
    "3. **Batch size** - we will also use these batch sizes to choose the optimum batch size, namely; `8, 16`. \n",
    "4. **Dropout value** - we will then use these dropout values to choose the optimum dropout value, namely; `0.5, 0.6, 0.7, 0.8, 0.9`\n",
    "\n",
    "> **Section one:** we will create common function and also create our base model, with the following initial parameters: **Batch size** `8`, **Optimizer** `Adam`, **Number of epochs** `5`, \n",
    "\n",
    "# **Import libraries that we will use**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "yDreG9NTlQ9a",
    "outputId": "c4adfe0f-9078-4e67-f5c4-4977f8c978b8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mounted at /content/drive\n"
     ]
    }
   ],
   "source": [
    "from google.colab import drive\n",
    "drive.mount('/content/drive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "kZ6pLmWClJUR"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "import cv2\n",
    "import random\n",
    "from keras_preprocessing.image import load_img\n",
    "from keras.preprocessing.image import img_to_array\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.layers import Dense, GlobalAveragePooling2D\n",
    "from tensorflow.keras.applications.vgg16 import VGG16 \n",
    "from tensorflow.keras.preprocessing import image \n",
    "from tensorflow.keras.applications.vgg16 import preprocess_input \n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from keras.layers import Input\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from tqdm import tqdm_notebook\n",
    "import time\n",
    "import tqdm.gui as tqdm\n",
    "import tqdm.notebook as tqdm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "AVS3DynBlJUS"
   },
   "source": [
    "# **Common functions**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sj9-x7mZlJUS"
   },
   "outputs": [],
   "source": [
    "\n",
    "def get_data_generator(train_data_path, \\\n",
    "                       val_data_path, \\\n",
    "                       targetsize, \\\n",
    "                       classmode, \\\n",
    "                       batchsize):\n",
    "    \"\"\"\n",
    "    This function is a data generator function for train, validation\n",
    "    Inputs\n",
    "        train_data_path   : train data path for the dataset \n",
    "        val_data_path     : validation data path for the dataset\n",
    "        targetsize        : target size for the generator to resize all images to, (224,224)\n",
    "        classmode         : class mode, 'categorical'\n",
    "        batchsize         : batch size\n",
    "    Outputs\n",
    "        train_generator   : generated train data\n",
    "        val_generator   : generated train data\n",
    "    \n",
    "    \"\"\"\n",
    "    train_datagen=ImageDataGenerator(preprocessing_function=preprocess_input)\n",
    "    val_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)\n",
    "\n",
    "    \n",
    "    train_generator=train_datagen.flow_from_directory(train_data_path, # this is where you specify the path to the main data folder\n",
    "                                                 target_size=targetsize,\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=batchsize,\n",
    "                                                 class_mode=classmode,\n",
    "                                                 shuffle=True)\n",
    "    \n",
    "    val_generator=val_datagen.flow_from_directory(val_data_path, # this is where you specify the path to the main data folder\n",
    "                                                 target_size=targetsize,\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=batchsize,\n",
    "                                                 class_mode=classmode,\n",
    "                                                 shuffle=True)\n",
    "    return train_generator,val_generator   \n",
    "\n",
    "\n",
    "\n",
    "def train_evaluate_the_model(train_generator, \\\n",
    "                            val_generator, \\\n",
    "                            optimizer, \\\n",
    "                            epochs, \\\n",
    "                            dropout_value, \\\n",
    "                            TheModel):\n",
    "    '''\n",
    "    train the model, do predictions, and do evaluation and return the accuracy of the model\n",
    "    Inputs\n",
    "        train_generator : generated train data\n",
    "        val_generator   : generated validation data\n",
    "        optimizer       : the optimizer method used to compile the model\n",
    "        epochs          : the epochs of the model use to fit the model\n",
    "        TheModel        : The pretrained model loaded,\n",
    "    Outputs\n",
    "        accuracy        : The accuracy of the model\n",
    "    '''\n",
    "    # load pretrained model and add layers on top of the model\n",
    "    x = TheModel.output\n",
    "    x = GlobalAveragePooling2D()(x)\n",
    "    x = Dense(1024,activation='relu')(x)\n",
    "    x = Dense(1024,activation='relu')(x)\n",
    "    x = Dense(512, activation='relu')(x)\n",
    "    x = tf.keras.layers.Dropout(dropout_value)(x)\n",
    "    preds = Dense(4,activation='softmax')(x)\n",
    "    model = Model(inputs = TheModel.input,outputs=preds)\n",
    "    \n",
    "    # freeze base layers for training\n",
    "    for layer in TheModel.layers:\n",
    "        layer.trainable = False\n",
    "    \n",
    "    # compile the model\n",
    "    model.compile(optimizer=optimizer,\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "    step_size_train=train_generator.n//train_generator.batch_size\n",
    "    # fit the model\n",
    "    r = model.fit_generator(generator=train_generator,\n",
    "                        validation_data=val_generator,\n",
    "                        steps_per_epoch=step_size_train,\n",
    "                        epochs=epochs)\n",
    "    # valuate the model\n",
    "    scores = model.evaluate(val_generator)\n",
    "    accuracy = scores[1]*100\n",
    "    \n",
    "    return accuracy, model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "9x6HPB3qvhH0"
   },
   "source": [
    "## **Explore the preprocess_input function**\n",
    "\n",
    "## The preprocess_input function will be used to preprocess the images so that they can be ready to be fed into the models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "MqpjtF9Gn7nk"
   },
   "outputs": [],
   "source": [
    "Gli_images=os.listdir(\"/content/drive/MyDrive/Capstone-Project/Training/Glioma/\")\n",
    "\n",
    "imgs=[]\n",
    "for i in range(len(Gli_images)):\n",
    "    img=load_img(Gli_images[i]) \n",
    "    imgs.append(img)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "nCkRZz0Xxccs",
    "outputId": "1a1bfb2f-4ed8-4a16-f3eb-998c72e17581"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maximum pixel value before preprocessing:  255.0\n"
     ]
    }
   ],
   "source": [
    "z=random.randint(1,len(imgs))\n",
    "im = img_to_array(imgs[z])\n",
    "print(\"Maximum pixel value before preprocessing: \",im.max())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "CEAHsvSlyDPU",
    "outputId": "fa6efa6b-7d8e-4762-fdd2-dc067b79b8dd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Maxium pixel value after pre-processing : 151.061\n"
     ]
    }
   ],
   "source": [
    "im = preprocess_input(im)\n",
    "print(\"Maxium pixel value after pre-processing :\",im.max())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 251
    },
    "id": "OAsEzoatoUcv",
    "outputId": "5c64e468-f9c3-4f3b-a62f-330849251795"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Preprocessed Image')"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "z=random.randint(1,len(imgs))\n",
    "im = img_to_array(imgs[z])\n",
    "im = np.expand_dims(im, axis=0)\n",
    "im = preprocess_input(im)\n",
    "\n",
    "plt.subplot(1,2,1)\n",
    "plt.imshow(imgs[z])\n",
    "plt.title(\"Original Image\")\n",
    "        \n",
    "        \n",
    "plt.subplot(1,2,2)\n",
    "im=np.reshape(im,(im.shape[1],im.shape[2],im.shape[3]))\n",
    "plt.imshow(im)\n",
    "plt.title(\"Preprocessed Image\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "m05ZMo-9lJUU"
   },
   "source": [
    "# **Load and generate data**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "B_RhpiZTlJUU",
    "outputId": "7d660e76-f838-4a29-b8de-2163b6991efe"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n"
     ]
    }
   ],
   "source": [
    "train_datagen=ImageDataGenerator(preprocessing_function=preprocess_input) #included in our dependencies\n",
    "val_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)\n",
    "\n",
    "'''\n",
    "Initial and constant parameters for the model\n",
    "initial parameters\n",
    "       batch_size    : the initial batch size we will create our base model with\n",
    "Constant parameters\n",
    "       target_size   : Constant image target size for the whole notebook which is (224,224)\n",
    "       class_mode    : class mode we will use for the whole notebook, which is 'categorical'\n",
    "'''\n",
    "# constant parameters\n",
    "targetsize_constant = (224,224)\n",
    "classmode_constant = 'categorical'\n",
    "\n",
    "# initial parameters\n",
    "batchsize_initial = 8\n",
    "epochs_initial = 4\n",
    "dropout_value_initial = 0.5\n",
    "optimizer_initial = 'Adam'\n",
    "\n",
    "# Data paths\n",
    "train_dataset_path = 'Dataset/Training/'\n",
    "validation_dataset_path = 'Dataset/Validation/'\n",
    "\n",
    "\n",
    "\n",
    "train_generator_initial = train_datagen.flow_from_directory(train_dataset_path, # this is where you specify the path to the main data folder\n",
    "                                                 target_size=targetsize_constant,\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=batchsize_initial,\n",
    "                                                 class_mode=classmode_constant,\n",
    "                                                 shuffle=True)\n",
    "val_generator_initial = val_datagen.flow_from_directory(validation_dataset_path, # this is where you specify the path to the main data folder\n",
    "                                                 target_size=targetsize_constant,\n",
    "                                                 color_mode='rgb',\n",
    "                                                 batch_size=batchsize_initial,\n",
    "                                                 class_mode=classmode_constant,\n",
    "                                                 shuffle=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2RdoIUTVlJUV",
    "outputId": "ee397b79-aa08-4875-be7c-a866d36ff190"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'glioma_tumor': 0, 'meningioma_tumor': 1, 'no_tumor': 2, 'pituitary_tumor': 3}"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# display classes found in the data\n",
    "train_generator_initial.class_indices"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UII7FOi2lJUW"
   },
   "source": [
    "#### Import the VGG16 base model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SHlUCzZ9lJUW"
   },
   "outputs": [],
   "source": [
    "image_size = [224,224] # choose image size\n",
    "# import the base model\n",
    "vgg = VGG16(input_shape= image_size+[3],weights='imagenet',include_top=False)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Xe43o0m4lJUX"
   },
   "source": [
    "# **Add layers on top of the base model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "tzpwp9e4lJUX"
   },
   "outputs": [],
   "source": [
    "x = vgg.output\n",
    "x = GlobalAveragePooling2D()(x)\n",
    "x = Dense(1024,activation='relu')(x)\n",
    "x = Dense(1024,activation='relu')(x)\n",
    "x = Dense(512, activation='relu')(x)\n",
    "preds = Dense(4,activation='softmax')(x)\n",
    "model = Model(inputs = vgg.input,outputs=preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "kp_FTdaTlJUY"
   },
   "outputs": [],
   "source": [
    "# freeze base layers for training\n",
    "for layer in vgg.layers:\n",
    "    layer.trainable = False"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-5eIdLQ3lJUY"
   },
   "source": [
    "# **Compile the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sZPxzF85lJUZ"
   },
   "outputs": [],
   "source": [
    "# initial optimizer\n",
    "optimizer_initial = 'Adam'\n",
    "\n",
    "# compile the model\n",
    "model.compile(optimizer=optimizer_initial,\n",
    "              loss='categorical_crossentropy',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lHjqbnpClJUZ"
   },
   "source": [
    "# **Fit the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QGWfL-4SlJUZ",
    "outputId": "10fa2a59-f648-4258-f73a-82a97e96ba71"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Student\\anaconda_3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py:1940: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "358/358 [==============================] - 932s 3s/step - loss: 0.7073 - accuracy: 0.7491 - val_loss: 0.4140 - val_accuracy: 0.8334\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 941s 3s/step - loss: 0.3818 - accuracy: 0.8557 - val_loss: 0.2514 - val_accuracy: 0.9056\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 949s 3s/step - loss: 0.2920 - accuracy: 0.8945 - val_loss: 0.2006 - val_accuracy: 0.9362\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 957s 3s/step - loss: 0.2659 - accuracy: 0.9001 - val_loss: 0.1971 - val_accuracy: 0.9373\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 968s 3s/step - loss: 0.2071 - accuracy: 0.9186 - val_loss: 0.2009 - val_accuracy: 0.9289\n"
     ]
    }
   ],
   "source": [
    "# initial parameters\n",
    "epochs_initial = 5\n",
    "step_size_train_initial=train_generator_initial.n//train_generator_initial.batch_size # computing the steps size per epoch\n",
    "\n",
    "tic = time.time()\n",
    "# fit the model\n",
    "r = model.fit_generator(generator=train_generator_initial,\n",
    "                        validation_data=val_generator_initial,\n",
    "                        steps_per_epoch=step_size_train_initial,\n",
    "                        epochs=epochs_initial)\n",
    "\n",
    "toc = time.time()\n",
    "print(\"Minutes taken = \" + str((toc-tic)/60.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Gnf2x8brlJUZ"
   },
   "source": [
    "# **Evaluate the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QVkHuQGmlJUa",
    "outputId": "2b257463-9ea7-42ae-b74b-04cafcd756e0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "359/359 [==============================] - 471s 1s/step - loss: 0.2009 - accuracy: 0.9289\n",
      "evaluate accuracy: 92.89%\n"
     ]
    }
   ],
   "source": [
    "# evaluation on validation data\n",
    "scores = model.evaluate(val_generator_initial)\n",
    "print(\"%s%s: %.2f%%\" % (\"evaluate \",model.metrics_names[1], scores[1]*100))\n",
    "\n",
    "#scores = model.evaluate_generator(val_generator)\n",
    "#print(\"%s%s: %.2f%%\" % (\"evaluate_generator \",model.metrics_names[1], scores[1]*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "mSAHyDLblJUa"
   },
   "source": [
    "# **Section two:** We will start fine tunning\n",
    "## **Tunning optimizers of the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "referenced_widgets": [
      "fe70544fe08a45af9046617e96f3479b"
     ]
    },
    "id": "fbYgb83elJUa",
    "outputId": "006a6911-2935-4150-e2c8-19e73d1aa9b5"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fe70544fe08a45af9046617e96f3479b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=5.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "358/358 [==============================] - 938s 3s/step - loss: 0.7257 - accuracy: 0.7530 - val_loss: 0.3804 - val_accuracy: 0.8460\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 933s 3s/step - loss: 0.3693 - accuracy: 0.8592 - val_loss: 0.3186 - val_accuracy: 0.8774\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 932s 3s/step - loss: 0.3001 - accuracy: 0.8823 - val_loss: 0.1630 - val_accuracy: 0.9373\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 932s 3s/step - loss: 0.2670 - accuracy: 0.9018 - val_loss: 0.1885 - val_accuracy: 0.9376\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 932s 3s/step - loss: 0.1999 - accuracy: 0.9217 - val_loss: 0.1324 - val_accuracy: 0.9505\n",
      "359/359 [==============================] - 467s 1s/step - loss: 0.1324 - accuracy: 0.9505\n",
      "Epoch 1/5\n",
      "358/358 [==============================] - 938s 3s/step - loss: 0.6824 - accuracy: 0.7568 - val_loss: 0.4280 - val_accuracy: 0.8073\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 937s 3s/step - loss: 0.3185 - accuracy: 0.8756 - val_loss: 0.2820 - val_accuracy: 0.8815\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 936s 3s/step - loss: 0.2280 - accuracy: 0.9175 - val_loss: 0.2479 - val_accuracy: 0.8958\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 935s 3s/step - loss: 0.1511 - accuracy: 0.9441 - val_loss: 0.3753 - val_accuracy: 0.8557\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 937s 3s/step - loss: 0.1181 - accuracy: 0.9532 - val_loss: 0.0713 - val_accuracy: 0.9812\n",
      "359/359 [==============================] - 468s 1s/step - loss: 0.0713 - accuracy: 0.9812\n",
      "Epoch 1/5\n",
      "358/358 [==============================] - 942s 3s/step - loss: 1.0042 - accuracy: 0.7034 - val_loss: 0.3995 - val_accuracy: 0.8251\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 942s 3s/step - loss: 0.5534 - accuracy: 0.8211 - val_loss: 0.4177 - val_accuracy: 0.8488\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 943s 3s/step - loss: 0.4907 - accuracy: 0.8581 - val_loss: 0.2933 - val_accuracy: 0.8930\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 942s 3s/step - loss: 0.4174 - accuracy: 0.8637 - val_loss: 0.3442 - val_accuracy: 0.8798\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 943s 3s/step - loss: 0.4177 - accuracy: 0.8819 - val_loss: 0.1657 - val_accuracy: 0.9394\n",
      "359/359 [==============================] - 470s 1s/step - loss: 0.1657 - accuracy: 0.9394\n",
      "Epoch 1/5\n",
      "358/358 [==============================] - 940s 3s/step - loss: 0.6181 - accuracy: 0.7655 - val_loss: 0.2996 - val_accuracy: 0.9063\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 939s 3s/step - loss: 0.3131 - accuracy: 0.8864 - val_loss: 0.2158 - val_accuracy: 0.9282\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 942s 3s/step - loss: 0.2288 - accuracy: 0.9224 - val_loss: 0.1662 - val_accuracy: 0.9551\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 944s 3s/step - loss: 0.1759 - accuracy: 0.9462 - val_loss: 0.1386 - val_accuracy: 0.9617\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 937s 3s/step - loss: 0.1452 - accuracy: 0.9546 - val_loss: 0.1093 - val_accuracy: 0.9774\n",
      "359/359 [==============================] - 470s 1s/step - loss: 0.1093 - accuracy: 0.9774\n",
      "Epoch 1/5\n",
      "358/358 [==============================] - 941s 3s/step - loss: 1.3341 - accuracy: 0.4050 - val_loss: 1.1604 - val_accuracy: 0.5073\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 939s 3s/step - loss: 1.0695 - accuracy: 0.5709 - val_loss: 0.9767 - val_accuracy: 0.6268\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 941s 3s/step - loss: 0.9166 - accuracy: 0.6565 - val_loss: 0.8528 - val_accuracy: 0.6990\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 941s 3s/step - loss: 0.8117 - accuracy: 0.7180 - val_loss: 0.7623 - val_accuracy: 0.7411\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 939s 3s/step - loss: 0.7334 - accuracy: 0.7505 - val_loss: 0.6952 - val_accuracy: 0.7620\n",
      "359/359 [==============================] - 471s 1s/step - loss: 0.6952 - accuracy: 0.7620\n",
      "\n",
      "Minutes taken = 430.23548024495443\n"
     ]
    },
    {
     "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>optimizer</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Adam</td>\n",
       "      <td>95.052266</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>SGD</td>\n",
       "      <td>98.118466</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>RMSprop</td>\n",
       "      <td>93.937284</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Adagrad</td>\n",
       "      <td>97.735190</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Adadelta</td>\n",
       "      <td>76.202089</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  optimizer   accuracy\n",
       "0      Adam  95.052266\n",
       "1       SGD  98.118466\n",
       "2   RMSprop  93.937284\n",
       "3   Adagrad  97.735190\n",
       "4  Adadelta  76.202089"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TheBaseModel = VGG16(input_shape= image_size+[3],weights='imagenet',include_top=False) # load a clean base model\n",
    "\n",
    "param_label = 'optimizer'\n",
    "param_list = ['Adam', 'SGD', 'RMSprop', 'Adagrad', 'Adadelta'] # ['Adam', 'SGD', 'RMSprop', 'Adagrad', 'Adadelta']\n",
    "\n",
    "accuracy_table = {param_label: [], 'accuracy': []}\n",
    "tic = time.time()\n",
    "for param in tqdm.tqdm_notebook(param_list):\n",
    "    # Train, predict and evaluate model\n",
    "    accuracy, _ = train_evaluate_the_model(train_generator_initial,val_generator_initial,param,epochs_initial,TheBaseModel) # replace epochs with 'epochs_initial'\n",
    "    \n",
    "    # Collect results\n",
    "    accuracy_table[param_label].append(param)\n",
    "    accuracy_table['accuracy'].append(accuracy)\n",
    "    \n",
    "accuracy_table = pd.DataFrame(accuracy_table) # convert the table to a dataframe\n",
    "toc = time.time()\n",
    "print(\"Minutes taken = \" + str((toc-tic)/60.0))\n",
    "accuracy_table # display the resullts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "jBiKTM5VlJUb",
    "outputId": "20c49e11-4a39-4477-c8bc-4ab7ceb3dfc2",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'accuracy)')"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from pylab import rcParams\n",
    "\n",
    "# plot results\n",
    "rcParams['figure.figsize'] = 10,8 # width 10, height 8\n",
    "\n",
    "ax = accuracy_table.plot(x='optimizer', y='accuracy',style='bx-', grid=True)\n",
    "#ax.set_xticklabels(param_list) # gave me an error when used\n",
    "ax.set_xlabel(\"Optimizer\")\n",
    "ax.set_ylabel(\"accuracy)\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "RYgT_ArulJUb",
    "outputId": "3a1361c1-0b83-423b-da80-a60236cfa2b5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "max Accuracy = 98.118\n",
      "optimum optimizer = SGD\n"
     ]
    }
   ],
   "source": [
    "# Get optimum value for param\n",
    "temp = accuracy_table[accuracy_table['accuracy'] == accuracy_table['accuracy'].max()]\n",
    "optimizer_opt = temp[param_label].values[0]\n",
    "print(\"max Accuracy = %0.3f\" % accuracy_table['accuracy'].max())\n",
    "print(\"optimum \" + param_label + \" = \" + str(optimizer_opt))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "F1FF3-0UlJUb"
   },
   "source": [
    "# **Tuning batch size and epochs of the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "referenced_widgets": [
      "ed5cf7921cc14861ad950dce28746219",
      "675da978549c49588f0f2e14843a4fa6",
      "8d3af7b3c60a4e2b93a4c0a0a080a515",
      "21bc45beaee04e61961b7c286f293f35",
      "a696e75f1a0c4936bc9a2c2e4825ac95"
     ]
    },
    "id": "5HnTCbS3lJUc",
    "outputId": "d13fe4a9-42ba-413f-ea31-9158ed2798d3"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ed5cf7921cc14861ad950dce28746219",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=4.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<ipython-input-15-d9397eb64c7a>:12: TqdmDeprecationWarning: This function will be removed in tqdm==5.0.0\n",
      "Please use `tqdm.notebook.tqdm` instead of `tqdm.tqdm_notebook`\n",
      "  for param2 in tqdm_notebook(param2_list):\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "675da978549c49588f0f2e14843a4fa6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=2.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Student\\anaconda_3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py:1940: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "358/358 [==============================] - 947s 3s/step - loss: 0.7642 - accuracy: 0.7537 - val_loss: 0.3878 - val_accuracy: 0.8613\n",
      "359/359 [==============================] - 472s 1s/step - loss: 0.3878 - accuracy: 0.8613\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "179/179 [==============================] - 937s 5s/step - loss: 0.8535 - accuracy: 0.7495 - val_loss: 0.3191 - val_accuracy: 0.8847\n",
      "180/180 [==============================] - 469s 3s/step - loss: 0.3191 - accuracy: 0.8847\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8d3af7b3c60a4e2b93a4c0a0a080a515",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=2.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/2\n",
      "358/358 [==============================] - 946s 3s/step - loss: 0.7618 - accuracy: 0.7338 - val_loss: 0.3760 - val_accuracy: 0.8707\n",
      "Epoch 2/2\n",
      "358/358 [==============================] - 945s 3s/step - loss: 0.3675 - accuracy: 0.8620 - val_loss: 0.2769 - val_accuracy: 0.9014\n",
      "359/359 [==============================] - 474s 1s/step - loss: 0.2769 - accuracy: 0.9014\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/2\n",
      "179/179 [==============================] - 944s 5s/step - loss: 0.9406 - accuracy: 0.7376 - val_loss: 0.2830 - val_accuracy: 0.8969\n",
      "Epoch 2/2\n",
      "179/179 [==============================] - 944s 5s/step - loss: 0.3432 - accuracy: 0.8732 - val_loss: 0.3030 - val_accuracy: 0.8850\n",
      "180/180 [==============================] - 473s 3s/step - loss: 0.3030 - accuracy: 0.8850\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "21bc45beaee04e61961b7c286f293f35",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=2.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/5\n",
      "358/358 [==============================] - 950s 3s/step - loss: 0.7569 - accuracy: 0.7481 - val_loss: 0.2794 - val_accuracy: 0.9017\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 949s 3s/step - loss: 0.3460 - accuracy: 0.8711 - val_loss: 0.2920 - val_accuracy: 0.8997\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 949s 3s/step - loss: 0.2844 - accuracy: 0.9008 - val_loss: 0.2848 - val_accuracy: 0.8704\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 948s 3s/step - loss: 0.2190 - accuracy: 0.9228 - val_loss: 0.1450 - val_accuracy: 0.9432\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 948s 3s/step - loss: 0.2176 - accuracy: 0.9203 - val_loss: 0.1211 - val_accuracy: 0.9669\n",
      "359/359 [==============================] - 475s 1s/step - loss: 0.1211 - accuracy: 0.9669\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/5\n",
      "179/179 [==============================] - 939s 5s/step - loss: 0.8689 - accuracy: 0.7446 - val_loss: 0.4010 - val_accuracy: 0.8446\n",
      "Epoch 2/5\n",
      "179/179 [==============================] - 937s 5s/step - loss: 0.3376 - accuracy: 0.8728 - val_loss: 0.2809 - val_accuracy: 0.8899\n",
      "Epoch 3/5\n",
      "179/179 [==============================] - 939s 5s/step - loss: 0.2615 - accuracy: 0.9036 - val_loss: 0.1840 - val_accuracy: 0.9331\n",
      "Epoch 4/5\n",
      "179/179 [==============================] - 936s 5s/step - loss: 0.2120 - accuracy: 0.9187 - val_loss: 0.1757 - val_accuracy: 0.9436\n",
      "Epoch 5/5\n",
      "179/179 [==============================] - 936s 5s/step - loss: 0.1796 - accuracy: 0.9359 - val_loss: 0.1008 - val_accuracy: 0.9648\n",
      "180/180 [==============================] - 469s 3s/step - loss: 0.1008 - accuracy: 0.9648\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a696e75f1a0c4936bc9a2c2e4825ac95",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=2.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/10\n",
      "358/358 [==============================] - 957s 3s/step - loss: 0.7263 - accuracy: 0.7582 - val_loss: 0.3282 - val_accuracy: 0.8686\n",
      "Epoch 2/10\n",
      "358/358 [==============================] - 957s 3s/step - loss: 0.3385 - accuracy: 0.8679 - val_loss: 0.2816 - val_accuracy: 0.8962\n",
      "Epoch 3/10\n",
      "358/358 [==============================] - 958s 3s/step - loss: 0.2682 - accuracy: 0.8920 - val_loss: 0.3231 - val_accuracy: 0.8760\n",
      "Epoch 4/10\n",
      "358/358 [==============================] - 957s 3s/step - loss: 0.2114 - accuracy: 0.9186 - val_loss: 0.1708 - val_accuracy: 0.9314\n",
      "Epoch 5/10\n",
      "358/358 [==============================] - 969s 3s/step - loss: 0.2160 - accuracy: 0.9252 - val_loss: 0.1075 - val_accuracy: 0.9648\n",
      "Epoch 6/10\n",
      "358/358 [==============================] - 958s 3s/step - loss: 0.1790 - accuracy: 0.9343 - val_loss: 0.1111 - val_accuracy: 0.9564\n",
      "Epoch 7/10\n",
      "358/358 [==============================] - 958s 3s/step - loss: 0.1517 - accuracy: 0.9476 - val_loss: 0.0898 - val_accuracy: 0.9714\n",
      "Epoch 8/10\n",
      "358/358 [==============================] - 979s 3s/step - loss: 0.1424 - accuracy: 0.9532 - val_loss: 0.1477 - val_accuracy: 0.9502\n",
      "Epoch 9/10\n",
      "358/358 [==============================] - 985s 3s/step - loss: 0.1111 - accuracy: 0.9584 - val_loss: 0.1283 - val_accuracy: 0.9505\n",
      "Epoch 10/10\n",
      "358/358 [==============================] - 957s 3s/step - loss: 0.0966 - accuracy: 0.9644 - val_loss: 0.0995 - val_accuracy: 0.9624\n",
      "359/359 [==============================] - 480s 1s/step - loss: 0.0995 - accuracy: 0.9624\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n",
      "Epoch 1/10\n",
      "179/179 [==============================] - 946s 5s/step - loss: 0.8054 - accuracy: 0.7582 - val_loss: 0.2741 - val_accuracy: 0.8972\n",
      "Epoch 2/10\n",
      "179/179 [==============================] - 946s 5s/step - loss: 0.3159 - accuracy: 0.8802 - val_loss: 0.3105 - val_accuracy: 0.8718\n",
      "Epoch 3/10\n",
      "179/179 [==============================] - 945s 5s/step - loss: 0.2627 - accuracy: 0.8984 - val_loss: 0.2315 - val_accuracy: 0.9174\n",
      "Epoch 4/10\n",
      "179/179 [==============================] - 947s 5s/step - loss: 0.2087 - accuracy: 0.9268 - val_loss: 0.1921 - val_accuracy: 0.9237\n",
      "Epoch 5/10\n",
      "179/179 [==============================] - 950s 5s/step - loss: 0.1852 - accuracy: 0.9306 - val_loss: 0.0797 - val_accuracy: 0.9735\n",
      "Epoch 6/10\n",
      "179/179 [==============================] - 947s 5s/step - loss: 0.1494 - accuracy: 0.9460 - val_loss: 0.0942 - val_accuracy: 0.9645\n",
      "Epoch 7/10\n",
      "179/179 [==============================] - 948s 5s/step - loss: 0.1438 - accuracy: 0.9481 - val_loss: 0.1677 - val_accuracy: 0.9296\n",
      "Epoch 8/10\n",
      "179/179 [==============================] - 947s 5s/step - loss: 0.1062 - accuracy: 0.9632 - val_loss: 0.0729 - val_accuracy: 0.9718\n",
      "Epoch 9/10\n",
      "179/179 [==============================] - 947s 5s/step - loss: 0.1146 - accuracy: 0.9611 - val_loss: 0.0858 - val_accuracy: 0.9725\n",
      "Epoch 10/10\n",
      "179/179 [==============================] - 945s 5s/step - loss: 0.0886 - accuracy: 0.9709 - val_loss: 0.1479 - val_accuracy: 0.9383\n",
      "180/180 [==============================] - 474s 3s/step - loss: 0.1479 - accuracy: 0.9383\n",
      "\n",
      "\n",
      "Minutes taken = 633.1540031512578\n"
     ]
    },
    {
     "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>epochs</th>\n",
       "      <th>batch_size</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>8</td>\n",
       "      <td>86.132401</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>16</td>\n",
       "      <td>88.466901</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>8</td>\n",
       "      <td>90.139371</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>16</td>\n",
       "      <td>88.501740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>8</td>\n",
       "      <td>96.689898</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>16</td>\n",
       "      <td>96.480834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>10</td>\n",
       "      <td>8</td>\n",
       "      <td>96.236932</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>10</td>\n",
       "      <td>16</td>\n",
       "      <td>93.832755</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   epochs  batch_size   accuracy\n",
       "0       1           8  86.132401\n",
       "1       1          16  88.466901\n",
       "2       2           8  90.139371\n",
       "3       2          16  88.501740\n",
       "4       5           8  96.689898\n",
       "5       5          16  96.480834\n",
       "6      10           8  96.236932\n",
       "7      10          16  93.832755"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TheBaseModel = VGG16(input_shape= image_size+[3],weights='imagenet',include_top=False) # load a clean base model\n",
    "\n",
    "param_label = 'epochs'\n",
    "param_list = [1,2,5,10] # [5, 10, 15, 20]\n",
    "\n",
    "param2_label = 'batch_size'\n",
    "param2_list = [8, 16] # [8, 16, 32] am using this batch sizes for now until we fix the accuracy issue, proper batch sizes to be use [8, 16, 32, 64, 128]\n",
    "\n",
    "accuracy_table = {param_label: [], param2_label: [], 'accuracy': []}\n",
    "tic = time.time()\n",
    "for param in tqdm.tqdm_notebook(param_list):\n",
    "    for param2 in tqdm_notebook(param2_list):\n",
    "        \n",
    "        # generate train and validation data\n",
    "        train_generator,val_generator, _ = get_data_generator(train_dataset_path,validation_dataset_path,test_dataset_path, targetsize_constant, classmode_constant, param2)\n",
    "        # train, predict and evaluate model\n",
    "        accuracy, _ = train_evaluate_the_model(train_generator, val_generator, optimizer_initial, param, TheBaseModel)\n",
    "        \n",
    "        # collect results\n",
    "        accuracy_table[param_label].append(param)\n",
    "        accuracy_table[param2_label].append(param2)\n",
    "        accuracy_table['accuracy'].append(accuracy)\n",
    "        \n",
    "accuracy_table = pd.DataFrame(accuracy_table) # convert the table to a dataframe\n",
    "toc = time.time()\n",
    "print(\"Minutes taken = \" + str((toc-tic)/60.0))\n",
    "accuracy_table # display the resullts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "lHDcky02lJUc",
    "outputId": "da1e26da-b2e5-4a9b-a745-dd55f7549084"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x213afd64af0>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot performance versus params\n",
    "rcParams['figure.figsize'] = 10, 8 # width 10, height 8\n",
    "temp = accuracy_table[accuracy_table[param2_label]==param2_list[0]]\n",
    "ax = temp.plot(x=param_label, y='accuracy', style='bs-', grid=True)\n",
    "legend_list = [param2_label + '_' + str(param2_list[0])]\n",
    "\n",
    "color_list = ['r', 'g', 'k', '0.75']\n",
    "for i in range(1,len(param2_list)):\n",
    "    temp = accuracy_table[accuracy_table[param2_label]==param2_list[i]]\n",
    "    ax = temp.plot(x=param_label, y='accuracy', color=color_list[i%len(color_list)], marker='s', grid=True, ax=ax)\n",
    "    legend_list.append(param2_label + '_' + str(param2_list[i]))\n",
    "\n",
    "ax.set_xlabel(param_label)\n",
    "ax.set_ylabel(\"accuracy\")\n",
    "plt.rcParams.update({'font.size': 14})\n",
    "plt.legend(legend_list, loc='center left', bbox_to_anchor=(1.0, 0.5)) # positions legend outside figure\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SBIy3Bu0lJUc",
    "outputId": "f0d53305-cd1d-43cf-e0cf-9d99ad9e47ec"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "max Accuracy = 96.690\n",
      "optimum epochs = 5\n",
      "optimum batch_size = 8\n"
     ]
    }
   ],
   "source": [
    "# Get optimum value for param and param2\n",
    "temp = accuracy_table[accuracy_table['accuracy'] == accuracy_table['accuracy'].max()]\n",
    "epochs_opt = temp[param_label].values[0]\n",
    "batch_size_opt = temp[param2_label].values[0]\n",
    "print(\"max Accuracy = %0.3f\" % accuracy_table['accuracy'].max())\n",
    "print(\"optimum \" + param_label + \" = \" + str(epochs_opt))\n",
    "print(\"optimum \" + param2_label + \" = \" + str(batch_size_opt))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UIwXb42OlJUc"
   },
   "source": [
    "# **Tunning dropout of the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "referenced_widgets": [
      "11f301c0f74b44a0b381ecab79bacbec"
     ]
    },
    "id": "hChDlbSalJUd",
    "outputId": "c5412adb-8b5e-4dd1-a790-9fb778a8c190"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "11f301c0f74b44a0b381ecab79bacbec",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=5.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Student\\anaconda_3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py:1940: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "358/358 [==============================] - 1136s 3s/step - loss: 0.8254 - accuracy: 0.7324 - val_loss: 0.4870 - val_accuracy: 0.8007\n",
      "Epoch 2/4\n",
      "358/358 [==============================] - 1132s 3s/step - loss: 0.4211 - accuracy: 0.8442 - val_loss: 0.3302 - val_accuracy: 0.8801\n",
      "Epoch 3/4\n",
      "358/358 [==============================] - 1076s 3s/step - loss: 0.3479 - accuracy: 0.8714 - val_loss: 0.2235 - val_accuracy: 0.9202\n",
      "Epoch 4/4\n",
      "358/358 [==============================] - 1080s 3s/step - loss: 0.2609 - accuracy: 0.9060 - val_loss: 0.1825 - val_accuracy: 0.9275\n",
      "359/359 [==============================] - 538s 1s/step - loss: 0.1825 - accuracy: 0.9275\n",
      "Epoch 1/4\n",
      "358/358 [==============================] - 1084s 3s/step - loss: 0.9356 - accuracy: 0.7261 - val_loss: 0.3973 - val_accuracy: 0.8526\n",
      "Epoch 2/4\n",
      "358/358 [==============================] - 1132s 3s/step - loss: 0.4669 - accuracy: 0.8274 - val_loss: 0.3752 - val_accuracy: 0.8666\n",
      "Epoch 3/4\n",
      "358/358 [==============================] - 1086s 3s/step - loss: 0.3197 - accuracy: 0.8829 - val_loss: 0.2591 - val_accuracy: 0.9045\n",
      "Epoch 4/4\n",
      "358/358 [==============================] - 1030s 3s/step - loss: 0.2868 - accuracy: 0.8945 - val_loss: 0.1707 - val_accuracy: 0.9310\n",
      "359/359 [==============================] - 515s 1s/step - loss: 0.1707 - accuracy: 0.9310\n",
      "Epoch 1/4\n",
      "358/358 [==============================] - 1052s 3s/step - loss: 1.0288 - accuracy: 0.6855 - val_loss: 0.3851 - val_accuracy: 0.8707\n",
      "Epoch 2/4\n",
      "358/358 [==============================] - 1030s 3s/step - loss: 0.4366 - accuracy: 0.8435 - val_loss: 0.4316 - val_accuracy: 0.8460\n",
      "Epoch 3/4\n",
      "358/358 [==============================] - 1029s 3s/step - loss: 0.3176 - accuracy: 0.8774 - val_loss: 0.2407 - val_accuracy: 0.9129\n",
      "Epoch 4/4\n",
      "358/358 [==============================] - 1031s 3s/step - loss: 0.3216 - accuracy: 0.8924 - val_loss: 0.2067 - val_accuracy: 0.9300\n",
      "359/359 [==============================] - 513s 1s/step - loss: 0.2067 - accuracy: 0.9300\n",
      "Epoch 1/4\n",
      "358/358 [==============================] - 1027s 3s/step - loss: 1.2053 - accuracy: 0.6548 - val_loss: 0.4078 - val_accuracy: 0.8341\n",
      "Epoch 2/4\n",
      "358/358 [==============================] - 1027s 3s/step - loss: 0.4809 - accuracy: 0.8323 - val_loss: 0.2800 - val_accuracy: 0.8990\n",
      "Epoch 3/4\n",
      "358/358 [==============================] - 1024s 3s/step - loss: 0.3574 - accuracy: 0.8739 - val_loss: 0.2445 - val_accuracy: 0.9045\n",
      "Epoch 4/4\n",
      "358/358 [==============================] - 1025s 3s/step - loss: 0.2982 - accuracy: 0.8973 - val_loss: 0.1775 - val_accuracy: 0.9355\n",
      "359/359 [==============================] - 512s 1s/step - loss: 0.1775 - accuracy: 0.9355\n",
      "Epoch 1/4\n",
      "358/358 [==============================] - 1044s 3s/step - loss: 1.5150 - accuracy: 0.5975 - val_loss: 0.5066 - val_accuracy: 0.8115\n",
      "Epoch 2/4\n",
      "358/358 [==============================] - 1050s 3s/step - loss: 0.6077 - accuracy: 0.7890 - val_loss: 0.4126 - val_accuracy: 0.8495\n",
      "Epoch 3/4\n",
      "358/358 [==============================] - 1048s 3s/step - loss: 0.4840 - accuracy: 0.8410 - val_loss: 0.2681 - val_accuracy: 0.9063\n",
      "Epoch 4/4\n",
      "358/358 [==============================] - 1041s 3s/step - loss: 0.4105 - accuracy: 0.8651 - val_loss: 0.2440 - val_accuracy: 0.9178\n",
      "359/359 [==============================] - 521s 1s/step - loss: 0.2440 - accuracy: 0.9178\n",
      "\n",
      "Minutes taken = 396.4226357460022\n"
     ]
    },
    {
     "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>dropout_prob</th>\n",
       "      <th>accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.5</td>\n",
       "      <td>92.752612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.6</td>\n",
       "      <td>93.101043</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.7</td>\n",
       "      <td>92.996514</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.8</td>\n",
       "      <td>93.554008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.9</td>\n",
       "      <td>91.777003</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   dropout_prob   accuracy\n",
       "0           0.5  92.752612\n",
       "1           0.6  93.101043\n",
       "2           0.7  92.996514\n",
       "3           0.8  93.554008\n",
       "4           0.9  91.777003"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "param_label = 'dropout_prob'\n",
    "param_list = [0.5, 0.6, 0.7, 0.8, 0.9]\n",
    "\n",
    "TheBaseModel = VGG16(input_shape= image_size+[3],weights='imagenet',include_top=False) # load a clean base model\n",
    "\n",
    "accuracy_table = {param_label: [], 'accuracy': []}\n",
    "tic = time.time()\n",
    "for param in tqdm.tqdm_notebook(param_list):\n",
    "    # Train, predict and evaluate model\n",
    "    accuracy, _ = train_evaluate_the_model(train_generator_initial,val_generator_initial,optimizer_initial,epochs_initial,param,TheBaseModel) # replace epochs with 'epochs_initial'\n",
    "    \n",
    "    # Collect results\n",
    "    accuracy_table[param_label].append(param)\n",
    "    accuracy_table['accuracy'].append(accuracy)\n",
    "    \n",
    "accuracy_table = pd.DataFrame(accuracy_table) # convert the table to a dataframe\n",
    "toc = time.time()\n",
    "print(\"Minutes taken = \" + str((toc-tic)/60.0))\n",
    "accuracy_table # display the resullts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "KlZJGKLllJUd",
    "outputId": "fc94c4b9-5e4d-4a72-d3c6-ee475c685148"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'accuracy)')"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from pylab import rcParams\n",
    "\n",
    "# plot results\n",
    "rcParams['figure.figsize'] = 10,8 # width 10, height 8\n",
    "\n",
    "ax = accuracy_table.plot(x='dropout_prob', y='accuracy',style='bx-', grid=True)\n",
    "#ax.set_xticklabels(param_list) # gave me an error when used\n",
    "ax.set_xlabel(\"dropout_prob\")\n",
    "ax.set_ylabel(\"accuracy)\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "OTKPVrvLlJUd",
    "outputId": "1c9e8a27-a6ae-48a6-f39c-d1b5822b2810"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "max Accuracy = 93.554\n",
      "optimum dropout_prob = 0.8\n"
     ]
    }
   ],
   "source": [
    "# Get optimum value for param\n",
    "temp = accuracy_table[accuracy_table['accuracy'] == accuracy_table['accuracy'].max()]\n",
    "dropout_opt = temp[param_label].values[0]\n",
    "print(\"max Accuracy = %0.3f\" % accuracy_table['accuracy'].max())\n",
    "print(\"optimum \" + param_label + \" = \" + str(dropout_opt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "qxyU9edalJUd"
   },
   "outputs": [],
   "source": [
    "\n",
    "dropout_opt = 0.8\n",
    "batch_size_opt = 8\n",
    "epochs_opt = 5\n",
    "optimizer_opt = 'SGD'\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Yyb1QGdslJUd"
   },
   "source": [
    "# **Collecting all optimum tuned parameters**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "HZ9MCmirlJUe",
    "outputId": "53cc17ae-c888-40e4-b1e1-e3ccea7e7d59"
   },
   "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>param</th>\n",
       "      <th>original</th>\n",
       "      <th>after_tuning</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>optimizer</td>\n",
       "      <td>Adam</td>\n",
       "      <td>SGD</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>epochs</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>batch_size</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>dropout_prob</td>\n",
       "      <td>0.5</td>\n",
       "      <td>0.8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          param original after_tuning\n",
       "0     optimizer     Adam          SGD\n",
       "1        epochs        4            5\n",
       "2    batch_size        8            8\n",
       "3  dropout_prob      0.5          0.8"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "d = {'param': ['optimizer', 'epochs', 'batch_size','dropout_prob'],\n",
    "     'original': [optimizer_initial, epochs_initial, batchsize_initial,dropout_value_initial],\n",
    "     'after_tuning': [optimizer_opt, epochs_opt, batch_size_opt,dropout_opt]}\n",
    "tuned_params = pd.DataFrame(d)\n",
    "tuned_params"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TQ7p6iJRlJUe"
   },
   "source": [
    "# **Creating the final model with optimum parameters**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QC1L4nCMlJUe",
    "outputId": "582498bf-718c-4ac2-c48f-58ad51d88c44"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2870 images belonging to 4 classes.\n",
      "Found 2870 images belonging to 4 classes.\n",
      "Found 394 images belonging to 4 classes.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Student\\anaconda_3\\lib\\site-packages\\tensorflow\\python\\keras\\engine\\training.py:1940: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.\n",
      "  warnings.warn('`Model.fit_generator` is deprecated and '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/5\n",
      "358/358 [==============================] - 1023s 3s/step - loss: 1.1318 - accuracy: 0.6338 - val_loss: 0.5083 - val_accuracy: 0.7958\n",
      "Epoch 2/5\n",
      "358/358 [==============================] - 1050s 3s/step - loss: 0.5398 - accuracy: 0.7883 - val_loss: 0.3276 - val_accuracy: 0.8808\n",
      "Epoch 3/5\n",
      "358/358 [==============================] - 1040s 3s/step - loss: 0.4479 - accuracy: 0.8239 - val_loss: 0.3719 - val_accuracy: 0.8505\n",
      "Epoch 4/5\n",
      "358/358 [==============================] - 1040s 3s/step - loss: 0.3779 - accuracy: 0.8641 - val_loss: 0.2510 - val_accuracy: 0.9122\n",
      "Epoch 5/5\n",
      "358/358 [==============================] - 1035s 3s/step - loss: 0.3171 - accuracy: 0.8788 - val_loss: 0.2036 - val_accuracy: 0.9230\n",
      "359/359 [==============================] - 520s 1s/step - loss: 0.2036 - accuracy: 0.9230\n"
     ]
    }
   ],
   "source": [
    "# get train and validation data generator\n",
    "train_generator_opt,val_generator_opt,test_generator_opt = get_data_generator(train_dataset_path,validation_dataset_path,test_dataset_path, targetsize_constant, classmode_constant, batch_size_opt)\n",
    "\n",
    "# load the base model\n",
    "The_base_Model = VGG16(input_shape= image_size+[3],weights='imagenet',include_top=False)\n",
    "\n",
    "# train and evaluate the model\n",
    "accuracy, New_model = train_evaluate_the_model(train_generator_opt, val_generator_opt, optimizer_opt, epochs_opt,dropout_opt, The_base_Model)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "76Yq-pLelJUe",
    "outputId": "c3054317-8102-415e-9d2d-9ae40cfecb5a"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "92.29965209960938"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bHoBCegflJUe"
   },
   "source": [
    "# **Save the final model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "qGiJBomJlJUe"
   },
   "outputs": [],
   "source": [
    "New_model.save('brain_tumor_classification_VGG16_test1309.h5') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8AI_l9g1lJUe"
   },
   "source": [
    "# ---------------------------------------- END OF NOTEBOOK -------------------------------------\n"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "FF VGG16 Model - Brain Tumor Classification + Model Creation and Fine Tunning .ipynb",
   "provenance": []
  },
  "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}