{"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_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# **CIFAR-10 using Convolutional Neural Network**\n---\n","metadata":{}},{"cell_type":"markdown","source":"**Install Required Libraries**","metadata":{}},{"cell_type":"code","source":"!pip install py7zr","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:23:37.432112Z","iopub.execute_input":"2022-07-19T06:23:37.432701Z","iopub.status.idle":"2022-07-19T06:23:50.688826Z","shell.execute_reply.started":"2022-07-19T06:23:37.432606Z","shell.execute_reply":"2022-07-19T06:23:50.688010Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Importing Required Libraries**","metadata":{}},{"cell_type":"code","source":"import tensorflow as tf\nimport shutil\nimport numpy as np\nimport cv2\nimport pandas as pd\nimport os\nfrom tensorflow import keras\nfrom keras.applications.xception import Xception, preprocess_input\nfrom keras.layers import Dense, MaxPool2D, Conv2D, Dropout, Flatten, GlobalAveragePooling2D, BatchNormalization, Activation, MaxPooling2D\nfrom keras.models import Sequential\nfrom keras.utils import np_utils\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom py7zr import unpack_7zarchive","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:23:50.690835Z","iopub.execute_input":"2022-07-19T06:23:50.691042Z","iopub.status.idle":"2022-07-19T06:23:56.783977Z","shell.execute_reply.started":"2022-07-19T06:23:50.691017Z","shell.execute_reply":"2022-07-19T06:23:56.783266Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Unpack the training dataset\nshutil.register_unpack_format('7zip',['.7z'],unpack_7zarchive)\nshutil.unpack_archive('../input/cifar-10/train.7z', '/kaggle/temp/')","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:23:56.785196Z","iopub.execute_input":"2022-07-19T06:23:56.785453Z","iopub.status.idle":"2022-07-19T06:24:44.450089Z","shell.execute_reply.started":"2022-07-19T06:23:56.785419Z","shell.execute_reply":"2022-07-19T06:24:44.449376Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_labels = pd.read_csv(\"../input/cifar-10/trainLabels.csv\", header=\"infer\")\n\nclasses = train_labels['label'].unique()\nprint(classes)","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:44.452138Z","iopub.execute_input":"2022-07-19T06:24:44.452409Z","iopub.status.idle":"2022-07-19T06:24:44.521774Z","shell.execute_reply.started":"2022-07-19T06:24:44.452375Z","shell.execute_reply":"2022-07-19T06:24:44.521063Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Divide the training set into training and validation dataset\nif not os.path.exists(\"/kaggle/temp/valid\"):\n    os.mkdir(\"/kaggle/temp/valid\")\n    \nparent_path_train = \"/kaggle/temp/train\"\nparent_path_valid = \"/kaggle/temp/valid\"\nparent_path_test = \"/kaggle/temp/test\"\n\nfor class1 in classes:\n    path_train = os.path.join(parent_path_train,class1)\n    if not os.path.exists(path_train):\n        os.mkdir(path_train)\n    path_valid = os.path.join(parent_path_valid,class1)\n    if not os.path.exists(path_valid):\n        os.mkdir(path_valid)\n        \nfor (int_ind,row) in train_labels.iterrows():\n    id = str(row[\"id\"])+\".png\"\n    source_path = os.path.join(parent_path_train,id)\n    \n    p=np.random.random()\n    if p<=0.8:\n        target_path = os.path.join(parent_path_train,row[\"label\"],id)\n        os.replace(source_path, target_path)\n    else:\n        target_path = os.path.join(parent_path_valid,row[\"label\"],id)\n        os.replace(source_path, target_path)","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:44.524523Z","iopub.execute_input":"2022-07-19T06:24:44.524743Z","iopub.status.idle":"2022-07-19T06:24:50.582172Z","shell.execute_reply.started":"2022-07-19T06:24:44.524712Z","shell.execute_reply":"2022-07-19T06:24:50.581339Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!ls /kaggle/temp/valid\n!ls /kaggle/temp/train","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:50.583355Z","iopub.execute_input":"2022-07-19T06:24:50.583801Z","iopub.status.idle":"2022-07-19T06:24:52.263161Z","shell.execute_reply.started":"2022-07-19T06:24:50.583764Z","shell.execute_reply":"2022-07-19T06:24:52.262235Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Performing Data Augmentation\ntrain_datagen = ImageDataGenerator(featurewise_center=False,\n                             samplewise_center=False,\n                             featurewise_std_normalization=False,\n                             samplewise_std_normalization=False,\n                             zca_whitening=False,\n                             rotation_range=10,\n                             zoom_range=0.1,\n                             width_shift_range=0.1,\n                             height_shift_range=0.1,\n                             horizontal_flip=False,\n                             vertical_flip=False,\n                             rescale=1./255)\nvalid_datagen = ImageDataGenerator()\n\ntrain_generator = train_datagen.flow_from_directory(directory='/kaggle/temp/train/', shuffle=True, target_size=(32,32),batch_size=128)\nvalid_generator = valid_datagen.flow_from_directory(directory='/kaggle/temp/valid/', shuffle=True, target_size=(32,32),batch_size=128)","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:52.265199Z","iopub.execute_input":"2022-07-19T06:24:52.265496Z","iopub.status.idle":"2022-07-19T06:24:55.077611Z","shell.execute_reply.started":"2022-07-19T06:24:52.265460Z","shell.execute_reply":"2022-07-19T06:24:55.076516Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Defining the model architecture\nmodel = Sequential()\nmodel.add(Conv2D(filters=32,kernel_size=(3,3),strides=(1,1),padding='valid',activation=None,use_bias=False,input_shape=(32,32,3)))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.3))\n\nmodel.add(Conv2D(filters=48, kernel_size=(3,3), strides=(1,1), padding='valid', activation=None, use_bias=False))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.3))\n\nmodel.add(Conv2D(filters=64, kernel_size=(3,3), strides=(1,1), padding='valid', activation=None, use_bias=False))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\n\nmodel.add(Conv2D(filters=80, kernel_size=(3,3), strides=(1,1), padding='valid', activation=None, use_bias=False))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.3))\n\nmodel.add(Conv2D(filters=96, kernel_size=(3,3), strides=(1,1), padding='valid', activation=None, use_bias=False))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\n\nmodel.add(Conv2D(filters=128, kernel_size=(3,3), strides=(1,1), padding='valid', activation=None, use_bias=False))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.3))\n\nmodel.add(Flatten())\nmodel.add(Dense(units=64))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.5))\n\nmodel.add(Dense(units=10))\nmodel.add(BatchNormalization())\nmodel.add(Activation('softmax'))\nmodel.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n\n# model.compile(tf.keras.optimizers.Nadam(\n#     learning_rate=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-07),loss=\"categorical_crossentropy\",metrics=[\"accuracy\"])","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:55.079049Z","iopub.execute_input":"2022-07-19T06:24:55.079488Z","iopub.status.idle":"2022-07-19T06:24:58.704107Z","shell.execute_reply.started":"2022-07-19T06:24:55.079447Z","shell.execute_reply":"2022-07-19T06:24:58.703364Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.fit(train_generator,epochs=30, validation_data=valid_generator,steps_per_epoch=train_generator.n//train_generator.batch_size,\n         validation_steps= valid_generator.n//valid_generator.batch_size,workers=8,use_multiprocessing=True)","metadata":{"execution":{"iopub.status.busy":"2022-07-19T06:24:58.705594Z","iopub.execute_input":"2022-07-19T06:24:58.706178Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Unpack the test dataset\nshutil.unpack_archive('/kaggle/input/cifar-10/test.7z','/kaggle/temp/test')\nshutil.unregister_unpack_format('7zip')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Perform preprocessing on test data\ntest_datagen = ImageDataGenerator(rescale=1./255)\n\ntest_gen = test_datagen.flow_from_directory(directory='/kaggle/temp/test',target_size=(32,32),batch_size=64,class_mode=None,shuffle=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_gen.reset()\npredictions_vecs = model.predict(test_gen)\n\npredictions_final = np.argmax(predictions_vecs, axis=1)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(type(train_generator.class_indices))\nprint(train_generator.class_indices)\n\nclasses = {value:key for (key,value) in train_generator.class_indices.items()}\nprint(classes)\n\npredicted_classes=np.empty(shape=300000,dtype=np.dtype('U20'))\n\nind=0\nfor i in predictions_final.tolist():\n    predicted_classes[ind]=classes[i]\n    ind=ind+1\n    \nfilenames_wo_ext = []\nfor fname in test_gen.filenames:\n    filenames_wo_ext.append(int(fname.split(sep=\"/\")[1].split(sep=\".\")[0])-1)\n\npredicted_classes_final = np.empty(shape=300000,dtype=np.dtype('U20'))\npredicted_classes_final[filenames_wo_ext]=predicted_classes","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sub = pd.read_csv('../input/cifar-10/sampleSubmission.csv',header='infer')\nsub.info()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sub['label'] = predicted_classes_final\nsub.to_csv('submission.csv',index=False)","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}