{"cells":[{"metadata":{},"cell_type":"markdown","source":"**Simple example of transfer learning from pretrained model using Keras.**\n* Loss: Focal loss\n* Metrics: f2_score"},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true},"cell_type":"code","source":"import os\nimport cv2\nimport matplotlib.pyplot as plt\nimport pandas as pd\nimport numpy as np\nimport json\nfrom keras.models import Sequential, Model\nfrom keras.layers import Dense, Flatten, Activation, Dropout, GlobalAveragePooling2D\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras import optimizers, applications\nfrom keras.callbacks import ModelCheckpoint, LearningRateScheduler, TensorBoard, EarlyStopping\nfrom keras import backend as K ","execution_count":1,"outputs":[{"output_type":"stream","text":"Using TensorFlow backend.\n","name":"stderr"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_df = pd.read_csv(\"../input/imet-2019-fgvc6/train.csv\")\ntrain_df[\"attribute_ids\"]=train_df[\"attribute_ids\"].apply(lambda x:x.split(\" \"))\ntrain_df[\"id\"]=train_df[\"id\"].apply(lambda x:x+\".png\")\ntrain_df.head()","execution_count":2,"outputs":[{"output_type":"execute_result","execution_count":2,"data":{"text/plain":" id attribute_ids\n0 1000483014d91860.png [147, 616, 813]\n1 1000fe2e667721fe.png [51, 616, 734, 813]\n2 1001614cb89646ee.png [776]\n3 10041eb49b297c08.png [51, 671, 698, 813, 1092]\n4 100501c227f8beea.png [13, 404, 492, 903, 1093]","text/html":"
\n\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
idattribute_ids
01000483014d91860.png[147, 616, 813]
11000fe2e667721fe.png[51, 616, 734, 813]
21001614cb89646ee.png[776]
310041eb49b297c08.png[51, 671, 698, 813, 1092]
4100501c227f8beea.png[13, 404, 492, 903, 1093]
\n
"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"label_df = pd.read_csv(\"../input/imet-2019-fgvc6/labels.csv\")\nprint(label_df.shape)\nlabel_df.head()","execution_count":3,"outputs":[{"output_type":"stream","text":"(1103, 2)\n","name":"stdout"},{"output_type":"execute_result","execution_count":3,"data":{"text/plain":" attribute_id attribute_name\n0 0 culture::abruzzi\n1 1 culture::achaemenid\n2 2 culture::aegean\n3 3 culture::afghan\n4 4 culture::after british","text/html":"
\n\n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n
attribute_idattribute_name
00culture::abruzzi
11culture::achaemenid
22culture::aegean
33culture::afghan
44culture::after british
\n
"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Example of images with tags\n\ni = 1\nplt.figure(figsize=[30,30])\nfor img_name in os.listdir(\"../input/imet-2019-fgvc6/train/\")[5:10]: \n img = cv2.imread(\"../input/imet-2019-fgvc6/train/%s\" % img_name)[...,[2, 1, 0]]\n plt.subplot(5, 1, i)\n plt.imshow(img)\n ids = train_df[train_df[\"id\"] == img_name][\"attribute_ids\"]\n title_val = []\n for tag_id in ids.values[0]:\n att_name = label_df[label_df['attribute_id'].astype(str) == tag_id]['attribute_name'].values[0]\n title_val.append(att_name)\n plt.title(title_val)\n i += 1\n \nplt.show()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"nb_classes = 1103\nbatch_size = 300\nimg_size = 75\nnb_epochs = 25","execution_count":4,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"lbls = list(map(str, range(nb_classes)))","execution_count":5,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\n\ntrain_datagen=ImageDataGenerator(\n rescale=1./255, \n validation_split=0.25,\n horizontal_flip = True, \n zoom_range = 0.3,\n width_shift_range = 0.3,\n height_shift_range=0.3\n )\n\ntrain_generator=train_datagen.flow_from_dataframe(\n dataframe=train_df,\n directory=\"../input/imet-2019-fgvc6/train\",\n x_col=\"id\",\n y_col=\"attribute_ids\",\n batch_size=batch_size,\n shuffle=True,\n class_mode=\"categorical\",\n classes=lbls,\n target_size=(img_size,img_size),\n subset='training')\n\nvalid_generator=train_datagen.flow_from_dataframe(\n dataframe=train_df,\n directory=\"../input/imet-2019-fgvc6/train\",\n x_col=\"id\",\n y_col=\"attribute_ids\",\n batch_size=batch_size,\n shuffle=True,\n class_mode=\"categorical\", \n classes=lbls,\n target_size=(img_size,img_size),\n subset='validation')","execution_count":6,"outputs":[{"output_type":"stream","text":"Found 81928 images belonging to 1103 classes.\nFound 27309 images belonging to 1103 classes.\nCPU times: user 5.04 s, sys: 7.2 s, total: 12.2 s\nWall time: 1min 5s\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Loss\n\ngamma = 2.0\nepsilon = K.epsilon()\ndef focal_loss(y_true, y_pred):\n pt = y_pred * y_true + (1-y_pred) * (1-y_true)\n pt = K.clip(pt, epsilon, 1-epsilon)\n CE = -K.log(pt)\n FL = K.pow(1-pt, gamma) * CE\n loss = K.sum(FL, axis=1)\n return loss","execution_count":7,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Metric\n\ndef f2_score(y_true, y_pred):\n beta = 2\n true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)), axis=1)\n predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)), axis=1)\n possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)), axis=1)\n \n precision = true_positives / (predicted_positives + K.epsilon())\n recall = true_positives / (possible_positives + K.epsilon())\n \n return K.mean(((1+beta**2)*precision*recall) / ((beta**2)*precision+recall+K.epsilon()))","execution_count":8,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model = applications.InceptionResNetV2(weights=None, \n include_top=False, \n input_shape=(img_size, img_size, 3))\nmodel.load_weights('../input/inceptionresnetv2/inception_resnet_v2_weights_tf_dim_ordering_tf_kernels_notop.h5')","execution_count":9,"outputs":[{"output_type":"stream","text":"WARNING:tensorflow:From /opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nColocations handled automatically by placer.\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.trainable = False","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Freeze some layers\n# for layer in model.layers[:-4]:\n# layer.trainable = False","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"#Adding custom layers \nx = model.output\nx = Flatten()(x)\nx = Dense(1024, activation=\"relu\")(x)\nx = Dropout(0.5)(x)\npredictions = Dense(nb_classes, activation=\"softmax\")(x)\nmodel_final = Model(input = model.input, output = predictions)\n\nmodel_final.compile(optimizers.rmsprop(lr=0.001, decay=1e-6),loss=focal_loss,metrics=[f2_score])","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-output":true,"trusted":true,"collapsed":true},"cell_type":"code","source":"# model_final.summary()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"# Callbacks\n\ncheckpoint = ModelCheckpoint(\"model_1.h5\", monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=False, mode='auto', period=1)\nearly = EarlyStopping(monitor='val_loss', min_delta=0, patience=5, verbose=1, mode='auto')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\nhistory = model_final.fit_generator(generator=train_generator, \n steps_per_epoch=500,\n validation_data=valid_generator, \n validation_steps=200,\n epochs=nb_epochs,\n callbacks = [checkpoint, early],\n max_queue_size=16,\n workers=2,\n use_multiprocessing=True,\n verbose=0)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"with open('history.json', 'w') as f:\n json.dump(history.history, f)\n\nhistory_df = pd.DataFrame(history.history)\nhistory_df[['loss', 'val_loss']].plot()\nhistory_df[['f2_score', 'val_f2_score']].plot()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"sam_sub_df = pd.read_csv('../input/imet-2019-fgvc6/sample_submission.csv')\nsam_sub_df[\"id\"]=sam_sub_df[\"id\"].apply(lambda x:x+\".png\")\nprint(sam_sub_df.shape)\nsam_sub_df.head()","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\ntest_datagen = ImageDataGenerator(rescale=1./255)\ntest_generator = test_datagen.flow_from_dataframe( \n dataframe=sam_sub_df,\n directory = \"../input/imet-2019-fgvc6/test\", \n x_col=\"id\",\n target_size = (img_size,img_size),\n batch_size = 1,\n shuffle = False,\n class_mode = None\n )","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\ntest_generator.reset()\npredict = model_final.predict_generator(test_generator, steps = len(test_generator.filenames))","execution_count":null,"outputs":[]},{"metadata":{"_kg_hide-output":true,"trusted":true},"cell_type":"code","source":"len(predict)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\nimport operator\npredicted_class_indices_3=[]\nfor i in range(len(predict)): \n d = {}\n for index, value in enumerate(predict[i]): \n if value > 0.03: \n d[index] = value \n sorted_d = sorted(d.items(), key=operator.itemgetter(1), reverse=True)\n \n # Take only first 10 items\n predicted_class_indices_3.append([i[0] for i in sorted_d[:10]])","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"%%time\npredictions_3=[]\n\nfor i in range(len(predicted_class_indices_3)):\n labels = (train_generator.class_indices)\n labels = dict((v,k) for k,v in labels.items())\n predictions = [labels[k] for k in predicted_class_indices_3[i]]\n predictions_3.append(predictions)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"predict_3 = []\nfor i in range(len(predictions_3)):\n str3 = \" \".join(predictions_3[i])\n predict_3.append(str3)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"filenames=test_generator.filenames\nresults=pd.DataFrame({\"id\":filenames,\n \"attribute_ids\":predict_3})\nresults['id'] = results['id'].map(lambda x: str(x)[:-4])\nresults.to_csv(\"submission.csv\",index=False)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"results.head()","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.4","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}