{"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 datetime\n\nfrom sklearn.model_selection import train_test_split\n\nfrom keras.utils.np_utils import to_categorical\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Flatten, Dropout, Conv2D, MaxPooling2D, Lambda, LeakyReLU\nfrom keras.layers.normalization import BatchNormalization\nfrom keras.optimizers import Adam\nfrom keras.callbacks import LearningRateScheduler","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"1, Train Data \n\ntrain.csv + Dig_MNIST.csv"},{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"train = pd.read_csv(\"/kaggle/input/Kannada-MNIST/train.csv\")\nDig = pd.read_csv(\"/kaggle/input/Kannada-MNIST/Dig-MNIST.csv\")\ntrain_data = train.append(Dig)\ny_train = np.array(train_data[\"label\"])\ny_train = to_categorical(y_train)\nx_train = train_data.drop(\"label\",axis=1).values\nx_train = x_train.reshape(x_train.shape[0], 28, 28, 1)\nx_train.shape, y_train.shape","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"2, Preprocessing Image"},{"metadata":{"trusted":true},"cell_type":"code","source":"X_train, X_valid, Y_train, Y_valid = train_test_split(x_train, y_train, test_size = 0.10, random_state=0) \n\ntrain_datagen = ImageDataGenerator(rescale = 1./255.,  #0~1内に\n                                   rotation_range = 10,  #回転範囲\n                                   width_shift_range = 0.25,  #水平シフト\n                                   height_shift_range = 0.25,  #垂直シフト\n                                   shear_range = 0.1,  #反時計回りのシアー角度\n                                   zoom_range = 0.25,  #ランダムにズームする範囲\n                                   horizontal_flip = False)  #水平方向に入力をランダムに反転\n\n\nvalid_datagen = ImageDataGenerator(rescale=1./255) ","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"3, Model"},{"metadata":{"trusted":true},"cell_type":"code","source":"model = Sequential()\nmodel.add(Conv2D(64, (3,3), padding='same', input_shape=(28, 28, 1)))\nmodel.add(BatchNormalization(momentum=0.5, epsilon=1e-5, gamma_initializer=\"uniform\"))\nmodel.add(LeakyReLU(alpha=0.1))\nmodel.add(Conv2D(64,  (3,3), padding='same'))\nmodel.add(BatchNormalization(momentum=0.1, epsilon=1e-5, gamma_initializer=\"uniform\"))\n          \nmodel.add(LeakyReLU(alpha=0.1))\nmodel.add(MaxPooling2D(2, 2))\nmodel.add(Dropout(0.2))\n          \nmodel.add(Conv2D(128, (3,3), padding='same'))\nmodel.add(BatchNormalization(momentum=0.2, epsilon=1e-5, gamma_initializer=\"uniform\"))\nmodel.add(LeakyReLU(alpha=0.1))\nmodel.add(Conv2D(128,  (3,3), padding='same'))\nmodel.add(BatchNormalization(momentum=0.1, epsilon=1e-5, gamma_initializer=\"uniform\"))       \n          \nmodel.add(LeakyReLU(alpha=0.1))\nmodel.add(MaxPooling2D(2, 2))\nmodel.add(Dropout(0.2))  \n          \nmodel.add(Conv2D(256, (3,3), padding='same'))\nmodel.add(BatchNormalization(momentum=0.2, epsilon=1e-5, gamma_initializer=\"uniform\"))\nmodel.add(LeakyReLU(alpha=0.1))\n          \nmodel.add(Conv2D(128, (3,3), padding='same'))\nmodel.add(BatchNormalization(momentum=0.1, epsilon=1e-5, gamma_initializer=\"uniform\"))\nmodel.add(LeakyReLU(alpha=0.1))\n          \nmodel.add(MaxPooling2D(2, 2))\nmodel.add(Dropout(0.2))\n          \nmodel.add(Flatten())        \nmodel.add(Dense(256,activation='relu',name='dense1'))\nmodel.add(LeakyReLU(alpha=0.1))          \nmodel.add(BatchNormalization())\nmodel.add(Dense(10,activation='softmax'))","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"4, Model Run"},{"metadata":{"trusted":true},"cell_type":"code","source":"\"\"\"initial_learningrate=1e-3\n\ndef lr_decay(epoch):\n    if epoch < 5:\n        return initial_learningrate\n    else:\n        return initial_learningrate * 0.99 ** epoch\"\"\"\n\ninitial_learningrate=2e-3\n\ndef lr_decay(epoch):\n    return initial_learningrate * 0.99 ** epoch\n    \nmodel.compile(loss=\"categorical_crossentropy\", optimizer=Adam(lr=initial_learningrate), metrics=[\"acc\"])\nmodel.fit_generator(train_datagen.flow(X_train, Y_train, batch_size=2048),\n                    steps_per_epoch=300,\n                    epochs=10,\n                    callbacks=[LearningRateScheduler(lr_decay,verbose=1)],\n                    validation_data=valid_datagen.flow(X_valid,Y_valid),\n                    validation_steps=50,  \n                    verbose=1)","execution_count":null,"outputs":[]},{"metadata":{},"cell_type":"markdown","source":"5, Create Submisson"},{"metadata":{"trusted":true},"cell_type":"code","source":"\nX_test = pd.read_csv(\"/kaggle/input/Kannada-MNIST/test.csv\")\nX_test = X_test.drop(\"id\", axis=1).values\nsubmission = pd.read_csv(\"/kaggle/input/Kannada-MNIST/sample_submission.csv\")\n\nX_test = X_test.reshape(X_test.shape[0], 28, 28, 1)\nX_test = X_test.astype(np.float32)/ 255\n\n\npredictions = model.predict_classes(X_test)\n\n\nsubmissions=pd.DataFrame({\"id\": submission[\"id\"], \"label\": predictions})\nsubmissions.to_csv(\"sUbmission.csv\", index=False)\n\nprint(\"FINISH\")","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":1}