{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n\n%matplotlib inline","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_dir = '../input/cassava-leaf-disease-classification/train_images/'\ntest_dir = '../input/cassava-leaf-disease-classification/test_images/'\n","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train_df = pd.read_csv('../input/cassava-leaf-disease-classification/train.csv')\nlabel_js = pd.read_json('../input/cassava-leaf-disease-classification/label_num_to_disease_map.json', \n                         orient='index')\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df['image'] = train_dir+train_df.image_id","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Disease sample photos"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.preprocessing import image\nfig, axes = plt.subplots(nrows=5, ncols=5, figsize=(20,20))\nfor row in np.arange(5):\n    for col in np.arange(5):        \n        cur_img_name = train_df[train_df.label==row].image.sample().values[0]\n        cur_img = image.load_img(cur_img_name)\n        cur_img = image.img_to_array(cur_img)\n        axes[row, col].imshow(cur_img/255., aspect='auto')  \n        axes[row, col].tick_params(axis='both', which='both', \n                                   bottom=False, top=False, \n                                   labelbottom=False, right=False, \n                                   left=False, labelleft=False)\nfor ax, label in zip(axes[:,0], label_js[0]):\n    ax.set_ylabel(label, rotation=90, size='large')\n\nplt.subplots_adjust(wspace=.05, hspace=.05)\nplt.show()\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Splitting dataset and encoding categorical target"},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.model_selection import train_test_split\nX_train, X_val, y_train, y_val = train_test_split(train_df['image'],train_df['label'], test_size=0.2)\n\nfrom sklearn.preprocessing import LabelBinarizer\nlb = LabelBinarizer()\nlb.fit(train_df.label)\n\ny_train_lb = lb.transform(y_train)\ny_val_lb = lb.transform(y_val)\n\nX_train_df = pd.DataFrame(X_train).reset_index().drop(labels='index', axis=1)\ny_train_df = pd.DataFrame(y_train_lb).add_prefix('label_')\n\nX_val_df = pd.DataFrame(X_val).reset_index().drop(labels='index', axis=1)\ny_val_df = pd.DataFrame(y_val_lb).add_prefix('label_')\n\ntrain = pd.concat([X_train_df, y_train_df], axis=1)\nvalidation = pd.concat([X_val_df, y_val_df], axis=1)\n\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Function for creating training and validation image generators"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.preprocessing.image import ImageDataGenerator\n\ndef create_image_generators(preprocess_input, target_image_size):\n    train_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)\n    val_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)\n\n    train_generator = train_datagen.flow_from_dataframe(\n        train,\n        x_col='image',\n        y_col=[f'label_{x}' for x in np.arange(5)],\n        target_size=target_image_size,\n        batch_size=32,\n        shuffle=True,\n        class_mode='raw')\n\n    validation_generator = val_datagen.flow_from_dataframe(\n        validation,\n        x_col='image',\n        y_col=[f'label_{x}' for x in np.arange(5)],\n        target_size=target_image_size,\n        shuffle=False,\n        batch_size=32,\n        class_mode='raw')\n    return train_generator, validation_generator","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"# ResNet50 model"},{"metadata":{},"cell_type":"markdown","source":"Image generators for training and validation"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.applications.resnet50 import preprocess_input as preprocess_input_rn\ntrain_generator_rn50, validation_generator_rn50 = create_image_generators(preprocess_input_rn, (224,224))","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Loading base model and previously saved weights"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.applications.resnet50 import ResNet50 \nbase_model_rn50 = ResNet50(input_shape=(224,224, 3),\n                        include_top=False, \n                        weights='imagenet')\n","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Adding layers for classification"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.layers import Flatten, Dense, GlobalAveragePooling2D, BatchNormalization, Activation, Dropout\nfrom keras.models import Model, Sequential\n\ndropout_dense_layer = 0.3\n\nmodel_rn50 = Sequential()\nmodel_rn50.add(base_model_rn50)\n    \nmodel_rn50.add(GlobalAveragePooling2D())\nmodel_rn50.add(Dense(128))\nmodel_rn50.add(BatchNormalization())\nmodel_rn50.add(Activation('relu'))\nmodel_rn50.add(Dense(32))\nmodel_rn50.add(BatchNormalization())\nmodel_rn50.add(Activation('relu'))\nmodel_rn50.add(Dropout(dropout_dense_layer))\n\nmodel_rn50.add(Dense(5, activation='softmax'))\n    \nmodel_rn50.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model_rn50.compile(optimizer='adam', \n              loss='categorical_crossentropy', \n              metrics=['accuracy'])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"EPOCHS = 5\nBATCH_SIZE = 32\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"history = model_rn50.fit_generator(generator=train_generator_rn50,\n                    validation_data=validation_generator_rn50,                    \n                    steps_per_epoch=len(train)//BATCH_SIZE,\n                    epochs=EPOCHS)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Predict labels"},{"metadata":{"trusted":true},"cell_type":"code","source":"import os\nimage_name_list = os.listdir(test_dir)\n\nlabel_list_rn50 = []\nlist_rn50 = []\n\n\n\nfor image_name in image_name_list:\n    img = image.load_img(test_dir+image_name, target_size=(224, 224))\n    model_input_img = preprocess_input_rn(np.expand_dims(img.copy(), axis=0)) \n    predicted_list_rn50 = model_rn50.predict(model_input_img)   \n    list_rn50.append(predicted_list_rn50)\n    predicted_label = np.argmax(predicted_list_rn50)\n    label_list_rn50.append(predicted_label)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"label_list_rn50","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}