{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true},"cell_type":"code","source":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        os.path.join(dirname, filename)\n\n# Any results you write to the current directory are saved as output.","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow import keras\nimport matplotlib.pyplot as plt\nimport pandas as pd\nimport os\nimport numpy as np\nimport cv2\nfrom PIL import Image\nimport shutil","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train = pd.read_csv(\"../input/aptos2019-blindness-detection/train.csv\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train.shape","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train['diagnosis'].unique()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train.diagnosis.hist()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def showbyserverity():\n    fig = plt.figure(figsize=(25, 16))\n    for class_id in sorted(train['diagnosis'].unique()): \n        for i, (idx, row) in enumerate(train.loc[train['diagnosis'] == class_id].sample(5, random_state=42).iterrows()):\n            ax = fig.add_subplot(5, 5, class_id * 5 + i + 1, xticks=[], yticks=[])\n            path=f\"../input/aptos2019-blindness-detection/train_images/{row['id_code']}.png\"\n            image = cv2.imread(path)\n            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n            image = cv2.resize(image, (256, 256))\n            image = cv2.addWeighted ( image,4, cv2.GaussianBlur( image , (0,0) , 50) ,-4 ,128)\n\n            plt.imshow(image, cmap = 'gist_gray')\n            ax.set_title('Label: %d-%d-%s' % (class_id, idx, row['id_code']) )","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"showbyserverity()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def label_img(name):\n    if name == 0 : \n        return np.array([1, 0, 0, 0, 0])\n    elif name == 1 : \n        return np.array([0, 1, 0, 0, 0])\n    elif name == 2 : \n        return np.array([0, 0, 1, 0, 0])\n    elif name == 3 : \n        return np.array([0, 0, 0, 1, 0])\n    else:\n        return np.array([0, 0, 0, 0, 1 ])","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Copied Individual ids into differnet folders "},{"metadata":{"trusted":true},"cell_type":"code","source":"for class_id in sorted(train['diagnosis'].unique()):\n    opath = f\"/output/kaggle/working/class_{class_id}\"\n    os.makedirs(opath)\n    for i, (idx, row) in enumerate(train.loc[train['diagnosis'] == class_id].iterrows()):\n        path=f\"../input/aptos2019-blindness-detection/train_images/{row['id_code']}.png\"\n        shutil.copy(path,opath)\n        \n    ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(os.listdir(\"/output/kaggle/working\"))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# shutil.rmtree(\"/output/kaggle/working\")","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Verifying the count of individual number of images in each id"},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(5):\n    opath = f\"/output/kaggle/working/class_{i}\"\n    list = os.listdir(opath) # dir is your directory path\n    print(len(list))\n   ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"!pip install Augmentor","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":" import Augmentor","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"def offline_augmentor(path, size, output_dir):\n    p = Augmentor.Pipeline(path,output_dir)\n    \n    p.rotate(probability=0.5, max_left_rotation=25, max_right_rotation=25)\n#     p.shear(probability=0.5, max_shear_left=16, max_shear_right=16)\n    p.flip_left_right(probability=0.5)\n    p.flip_top_bottom(probability=0.5)\n    p.zoom(probability=0.5, min_factor=0.75, max_factor=1.3)\n    p.crop_random(probability=0.5, percentage_area=0.9)\n#     p.random_brightness(probability=0.5, max_factor=1.2, min_factor=0.4)\n#     p.random_color(probability=0.5, max_factor=0.8, min_factor=0.3)\n#     p.random_contrast(probability=0.5, max_factor=0.8, min_factor=0.3)\n    p.resize(probability=1.0, width=512, height=512)\n    p.sample(size)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path = '/output/kaggle/working/class_1'\nsize = 1400\noutput_dir='/output/kaggle/working/class_1/output'\n\noffline_augmentor(path, size, output_dir)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path = '/output/kaggle/working/class_2'\nsize = 800\noutput_dir='/output/kaggle/working/class_2/output'\n\noffline_augmentor(path, size, output_dir)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path = '/output/kaggle/working/class_3'\nsize = 1600\noutput_dir='/output/kaggle/working/class_3/output'\n\noffline_augmentor(path, size, output_dir)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path = '/output/kaggle/working/class_4'\nsize = 1500\noutput_dir='/output/kaggle/working/class_4/output'\n\noffline_augmentor(path, size, output_dir)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(os.listdir(\"/output/kaggle/working/class_4/output\"))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path=f\"/output/kaggle/working/class_4/output/class_4_original_e019b3e0f33d.png_68e673f0-8f10-4e21-bbd4-7b88e4aa10ed.png\"\nimage = cv2.imread(path)\nplt.imshow(image)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"path=f\"../input/aptos2019-blindness-detection/train_images/e019b3e0f33d.png\"\nimage = cv2.imread(path)\nplt.imshow(image)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_image=[]\ntrain_label=[]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(0,5):\n    path = f\"/output/kaggle/working/class_{i}\"\n    for filename in os.listdir(path):\n        if filename.endswith(\".jpg\") or filename.endswith(\".png\"):\n            path_img = os.path.join(path,filename)\n            image = cv2.imread(path_img)\n            label = label_img(i)\n            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n            image = cv2.resize(image, (256, 256))\n            image = cv2.addWeighted ( image,4, cv2.GaussianBlur( image , (0,0) , 50) ,-4 ,128)\n            train_image.append(np.array(image))\n            train_label.append(label)\n    if i!=0:\n        path = f\"/output/kaggle/working/class_{i}/output\"\n        for filename in os.listdir(path):\n            if filename.endswith(\".jpg\") or filename.endswith(\".png\"):\n                path_img = os.path.join(path,filename)\n                image = cv2.imread(path_img)\n                label = label_img(i)\n                image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n                image = cv2.resize(image, (256, 256))\n                image = cv2.addWeighted ( image,4, cv2.GaussianBlur( image , (0,0) , 50) ,-4 ,128)\n                train_image.append(np.array(image))\n                train_label.append(label)    \n    \n        ","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_image[255].shape","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.model_selection import train_test_split\nxTrain, xTest, yTrain, yTest = train_test_split(train_image, train_label, test_size = 0.25, random_state = 42)\n\nxTrain, xVal, yTrain, yVal = train_test_split(xTrain, yTrain, test_size=0.15, random_state=42)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"type(xTrain)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"xTrain = np.array(xTrain)\nxTest = np.array(xTest)\n\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"yTrain = np.array(yTrain)\nyTest = np.array(yTest)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"xVal = np.array(xVal)\nyVal = np.array(yVal)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"import matplotlib.pyplot as plt\nplt.imshow(train_image[123], cmap = 'gist_gray')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"sumy =[]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(len(yTrain)):\n    index = np.argmax(yTrain[i])\n    if index == 0:        \n        sumy.append(0)\n    elif index == 1: \n        sumy.append(1)\n    elif index == 2: \n        sumy.append(2)\n    elif index == 3:\n        sumy.append(3)\n    elif index == 4:\n        sumy.append(4)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from collections import Counter","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"Counter(sumy)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.models import Sequential\nfrom keras.layers import Dense, Dropout, Activation, Flatten\nfrom keras.layers import Conv2D, MaxPooling2D, BatchNormalization\nmodel = Sequential()\nmodel.add(Conv2D(32, kernel_size = (3, 3), activation='relu', input_shape=(256, 256, 3)))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(64, kernel_size=(3,3), activation='relu'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(64, kernel_size=(3,3), activation='relu'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(96, kernel_size=(3,3), activation='relu'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(32, kernel_size=(3,3), activation='relu'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.2))\nmodel.add(Flatten())\nmodel.add(Dense(128, activation='relu'))\n# model.add(Dropout(0.3))\nmodel.add(Dense(5, activation = 'softmax'))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for layer in model.layers:\n    print(layer.output_shape)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.optimizers import SGD\nfrom keras import metrics\nsgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)\nmodel.compile(loss='categorical_crossentropy', metrics = [metrics.categorical_accuracy],optimizer='adam')\n\n","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.callbacks import ModelCheckpoint","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"checkpointer = ModelCheckpoint(filepath=\"best_weights.hdf5\", monitor = 'val_categorical_accuracy',verbose=1, save_best_only=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"history = model.fit(xTrain, yTrain, batch_size=32, epochs=40,callbacks=[checkpointer],validation_data=(xVal,yVal))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.load_weights('best_weights.hdf5')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.save('model_1.h5')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"acc = history.history['categorical_accuracy']\nloss = history.history['loss']\n\n\nepochs = range(1, len(acc) + 1)\n\nplt.plot(epochs, acc, 'bo', label='Training acc')\nplt.title('Training accuracy')\nplt.legend()\n\nplt.figure()\n\nplt.plot(epochs, loss, 'bo', label='Training loss')\nplt.title('Training loss')\nplt.legend()\n\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"yPred = model.predict(xTest)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"result = []","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(len(yPred)):\n    index = np.argmax(yPred[i])\n    if index == 0:        #According to one hot encoding above, 0 is Coronal, 1 is Horizontal and 2 is Sagittal.\n        result.append(0)\n    elif index == 1: \n        result.append(1)\n    elif index == 2: \n        result.append(2)\n    elif index == 3:\n        result.append(3)\n    elif index == 4:\n        result.append(4)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"result_test=[]","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"for i in range(len(yTest)):\n    index = np.argmax(yTest[i])\n    if index == 0:        #According to one hot encoding above, 0 is Coronal, 1 is Horizontal and 2 is Sagittal.\n        result_test.append(0)\n    elif index == 1: \n        result_test.append(1)\n    elif index == 2: \n        result_test.append(2)\n    elif index == 3:\n        result_test.append(3)\n    elif index == 4:\n        result_test.append(4)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from sklearn.metrics import accuracy_score,confusion_matrix,f1_score,recall_score,precision_score,precision_recall_fscore_support","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"f1_score(result,result_test,average='macro')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"confusion_matrix(result,result_test)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"accuracy_score(result,result_test)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"loss, accuracy = model.evaluate(xTest,yTest, batch_size=32)\nprint(loss, accuracy)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"Model with the best hyperparameters from previous book"},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.models import Sequential\nfrom keras.layers import Dense, Dropout, Activation, Flatten\nfrom keras.layers import Conv2D, MaxPooling2D, BatchNormalization\nmodel = Sequential()\nmodel.add(Conv2D(32, kernel_size = (3, 3), activation='relu',kernel_initializer='uniform',input_shape=(256, 256, 3)))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(64, kernel_size=(3,3), activation='relu',kernel_initializer='uniform'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(64, kernel_size=(3,3), activation='relu',kernel_initializer='uniform'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(96, kernel_size=(3,3), activation='relu',kernel_initializer='uniform'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(32, kernel_size=(3,3), activation='relu',kernel_initializer='uniform'))\nmodel.add(MaxPooling2D(pool_size=(2,2)))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.2))\nmodel.add(Flatten())\nmodel.add(Dense(128, activation='relu',kernel_initializer='uniform'))\n# model.add(Dropout(0.3))\nmodel.add(Dense(5, activation = 'softmax',kernel_initializer='uniform'))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.callbacks import ModelCheckpoint","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras import metrics\n\nmodel.compile(loss='categorical_crossentropy',metrics=[metrics.categorical_accuracy],optimizer='rmsprop')\ncheckpointer = ModelCheckpoint(filepath='best_weights.hdf5',monitor='val_categorical_accuracy',verbose=1,save_best_only=True)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"history = model.fit(xTrain, yTrain, batch_size=32, epochs=40,callbacks=[checkpointer],validation_data=(xVal,yVal))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.load_weights('best_weights.hdf5')\nmodel.save('model_2.h5')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"acc = history.history['categorical_accuracy']\nloss = history.history['loss']\n\n\nepochs = range(1, len(acc) + 1)\n\nplt.plot(epochs, acc, 'bo', label='Training acc')\nplt.title('Training accuracy')\nplt.legend()\n\nplt.figure()\n\nplt.plot(epochs, loss, 'bo', label='Training loss')\nplt.title('Training loss')\nplt.legend()\n\nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"yPred = model.predict(xTest)\nresult = []\nfor i in range(len(yPred)):\n    index = np.argmax(yPred[i])\n    if index == 0:        #According to one hot encoding above, 0 is Coronal, 1 is Horizontal and 2 is Sagittal.\n        result.append(0)\n    elif index == 1: \n        result.append(1)\n    elif index == 2: \n        result.append(2)\n    elif index == 3:\n        result.append(3)\n    elif index == 4:\n        result.append(4)\nresult_test=[]\nfor i in range(len(yTest)):\n    index = np.argmax(yTest[i])\n    if index == 0:        #According to one hot encoding above, 0 is Coronal, 1 is Horizontal and 2 is Sagittal.\n        result_test.append(0)\n    elif index == 1: \n        result_test.append(1)\n    elif index == 2: \n        result_test.append(2)\n    elif index == 3:\n        result_test.append(3)\n    elif index == 4:\n        result_test.append(4)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"recall_score(result,result_test,average='macro')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"confusion_matrix(result,result_test)","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}