{"cells":[{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"import os\nfrom glob import glob\nimport random\nimport time\nimport tensorflow\nimport datetime\nos.environ['KERAS_BACKEND'] = 'tensorflow'\nos.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' # 3 = INFO, WARNING, and ERROR messages are not printed\n\nfrom tqdm import tqdm\n\nimport numpy as np\nimport pandas as pd\nfrom IPython.display import FileLink\nimport matplotlib.pyplot as plt\nimport warnings\nwarnings.filterwarnings('ignore')\nimport seaborn as sns \n%matplotlib inline\nfrom IPython.display import display, Image\nimport matplotlib.image as mpimg\nimport cv2\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.datasets import load_files       \nfrom keras.utils import np_utils\nfrom sklearn.utils import shuffle\nfrom sklearn.metrics import log_loss\n\nfrom keras.models import Sequential, Model\nfrom keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, BatchNormalization, GlobalAveragePooling2D\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras.preprocessing import image\nfrom keras.callbacks import ModelCheckpoint, EarlyStopping\nfrom keras.applications.vgg16 import VGG16\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"dataset = pd.read_csv('../input/state-farm-distracted-driver-detection/driver_imgs_list.csv')\ndataset.head(5)\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"by_drivers = dataset.groupby('subject')\nunique_drivers = by_drivers.groups.keys()\nprint(unique_drivers)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Load the dataset previously downloaded from Kaggle\nNUMBER_CLASSES = 10\n# Color type: 1 - grey, 3 - rgb\n\ndef get_cv2_image(path, img_rows, img_cols, color_type=3):\n    # Loading as Grayscale image\n    if color_type == 1:\n        img = cv2.imread(path, cv2.IMREAD_GRAYSCALE)\n    elif color_type == 3:\n        img = cv2.imread(path, cv2.IMREAD_COLOR)\n    # Reduce size\n    img = cv2.resize(img, (img_rows, img_cols)) \n    return img\n\n# Training\ndef load_train(img_rows, img_cols, color_type=3):\n    start_time = time.time()\n    train_images = [] \n    train_labels = []\n    # Loop over the training folder \n    for classed in tqdm(range(NUMBER_CLASSES)):\n        print('Loading directory c{}'.format(classed))\n        files = glob(os.path.join('../input/state-farm-distracted-driver-detection/imgs/train' , 'c' + str(classed), '*.jpg'))\n        for file in files:\n            img = get_cv2_image(file, img_rows, img_cols, color_type)\n            train_images.append(img)\n            train_labels.append(classed)\n    print(\"Data Loaded in {} second\".format(time.time() - start_time))\n    return train_images, train_labels \n\ndef read_and_normalize_train_data(img_rows, img_cols, color_type):\n    X, labels = load_train(img_rows, img_cols, color_type)\n    y = np_utils.to_categorical(labels, 10)\n    x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n    \n    x_train = np.array(x_train, dtype=np.uint8).reshape(-1,img_rows,img_cols,color_type)\n    x_test = np.array(x_test, dtype=np.uint8).reshape(-1,img_rows,img_cols,color_type)\n    \n    return x_train, x_test, y_train, y_test\n\n# Validation\ndef load_test(size=200000, img_rows=64, img_cols=64, color_type=3):\n    path = os.path.join('../input/state-farm-distracted-driver-detection/imgs/test' ,'c' + str(classed), '*.jpg')\n    files = sorted(glob(path))\n    X_test, X_test_id = [], []\n    total = 0\n    files_size = len(files)\n    for file in tqdm(files):\n        if total >= size or total >= files_size:\n            break\n        file_base = os.path.basename(file)\n        img = get_cv2_image(file, img_rows, img_cols, color_type)\n        X_test.append(img)\n        X_test_id.append(file_base)\n        total += 1\n    return X_test, X_test_id\n\ndef read_and_normalize_sampled_test_data(size, img_rows, img_cols, color_type=3):\n    test_data, test_ids = load_test(size, img_rows, img_cols, color_type)\n    \n    test_data = np.array(test_data, dtype=np.uint8)\n    test_data = test_data.reshape(-1,img_rows,img_cols,color_type)\n    \n    return test_data, test_ids\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"img_rows = 224\nimg_cols = 224\ncolor_type = 1","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"x_train, x_test, y_train, y_test = read_and_normalize_train_data(img_rows, img_cols, color_type)\nprint('Train shape:', x_train.shape)\nprint(x_train.shape[0], 'train samples')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Statistics\n# Load the list of names\nnames = [item[17:19] for item in sorted(glob(\"../input/state-farm-distracted-driver-detection/imgs/train/*/\"))]\ntest_files_size = len(np.array(glob(os.path.join('../input/state-farm-distracted-driver-detection/imgs', 'test', '*.jpg'))))\nx_train_size = len(x_train)\ncategories_size = len(names)\nx_test_size = len(x_test)\nprint('There are %s total images.\\n' % (test_files_size + x_train_size + x_test_size))\nprint('There are %d training images.' % x_train_size)\nprint('There are %d total training categories.' % categories_size)\nprint('There are %d validation images.' % x_test_size)\nprint('There are %d test images.'% test_files_size)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"activity_map = {'c0': 'Safe driving', \n                'c1': 'Texting - right', \n                'c2': 'Talking on the phone - right', \n                'c3': 'Texting - left', \n                'c4': 'Talking on the phone - left', \n                'c5': 'Operating the radio', \n                'c6': 'Drinking', \n                'c7': 'Reaching behind', \n                'c8': 'Hair and makeup', \n                'c9': 'Talking to passenger'}","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Plot figure size\nplt.figure(figsize = (10,10))\n# Count the number of images per category\nsns.countplot(x = 'classname', data = dataset)\n# Change the Axis names\nplt.ylabel('Count')\nplt.title('Categories Distribution')\n# Show plot\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"batch_size = 40\nnb_epoch = 30","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"!rm -f saved_models/weights_best_vanilla2.hdf5","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"models_dir = \"saved_models\"\nif not os.path.exists(models_dir):\n    os.makedirs(models_dir)\n    \ncheckpointer = ModelCheckpoint(filepath='saved_models/weights_best_vanilla2.hdf5', \n                               monitor='val_loss', mode='min',\n                               verbose=1, save_best_only=True)\nes = EarlyStopping(monitor='val_acc', mode='max', verbose=1, patience=5)\ncallbacks = [checkpointer, es]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def create_model_v2():\n    # Optimised Vanilla CNN model\n    model = Sequential()\n\n    ## CNN 1\n    model.add(Conv2D(32,(3,3),activation='relu',input_shape=(img_rows, img_cols, color_type)))\n    model.add(BatchNormalization())\n    model.add(Conv2D(32,(3,3),activation='relu',padding='same'))\n    model.add(BatchNormalization(axis = 3))\n    model.add(MaxPooling2D(pool_size=(2,2),padding='same'))\n    model.add(Dropout(0.3))\n\n    ## CNN 2\n    model.add(Conv2D(64,(3,3),activation='relu',padding='same'))\n    model.add(BatchNormalization())\n    model.add(Conv2D(64,(3,3),activation='relu',padding='same'))\n    model.add(BatchNormalization(axis = 3))\n    model.add(MaxPooling2D(pool_size=(2,2),padding='same'))\n    model.add(Dropout(0.3))\n\n    ## CNN 3\n    model.add(Conv2D(128,(3,3),activation='relu',padding='same'))\n    model.add(BatchNormalization())\n    model.add(Conv2D(128,(3,3),activation='relu',padding='same'))\n    model.add(BatchNormalization(axis = 3))\n    model.add(MaxPooling2D(pool_size=(2,2),padding='same'))\n    model.add(Dropout(0.5))\n\n    ## Output\n    model.add(Flatten())\n    model.add(Dense(512,activation='relu'))\n    model.add(BatchNormalization())\n    model.add(Dropout(0.5))\n    model.add(Dense(128,activation='relu'))\n    model.add(Dropout(0.25))\n    model.add(Dense(10,activation='softmax'))\n\n    return model","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model_v2 = create_model_v2()\n\n# More details about the layers\nmodel_v2.summary()\n\n# Compiling the model\nmodel_v2.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Training the Vanilla Model\nhistory_v2 = model_v2.fit(x_train, y_train, \n          validation_data=(x_test, y_test),\n          callbacks=callbacks,\n          epochs=nb_epoch, batch_size=batch_size, verbose=1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def plot_train_history(history):\n    # Summarize history for accuracy\n    plt.plot(history.history['accuracy'])\n    plt.plot(history.history['val_accuracy'])\n    plt.title('Model accuracy')\n    plt.ylabel('accuracy')\n    plt.xlabel('epoch')\n    plt.legend(['train', 'test'], loc='upper left')\n    plt.show()\n\n    # Summarize history for loss\n    plt.plot(history.history['loss'])\n    plt.plot(history.history['val_loss'])\n    plt.title('Model loss')\n    plt.ylabel('loss')\n    plt.xlabel('epoch')\n    plt.legend(['train', 'test'], loc='upper left')\n    plt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"plot_train_history(history_v2)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"score = model_v2.evaluate(x_test, y_test, verbose=1)\nprint('Score: ', score)\n\ny_pred = model_v2.predict(x_test, batch_size=batch_size, verbose=1)\nscore = log_loss(y_test, y_pred)\nprint('Score log loss:', score)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def plot_test_class(model, test_files, image_number, color_type=1):\n    img_brute = test_files[image_number]\n    img_brute = cv2.resize(img_brute,(img_rows,img_cols))\n    plt.imshow(img_brute, cmap='gray')\n\n    new_img = img_brute.reshape(-1,img_rows,img_cols,color_type)\n\n    y_prediction = model.predict(new_img, batch_size=batch_size, verbose=1)\n    print('Y prediction: {}'.format(y_prediction))\n    print('Predicted: {}'.format(activity_map.get('c{}'.format(np.argmax(y_prediction)))))\n    \n    plt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model_v2.save('my_model.h5')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"    import os\n    os.chdir('/kaggle/working')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from IPython.display import FileLink\nFileLink('my_model.h5')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat":4,"nbformat_minor":4}