{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":5441,"databundleVersionId":38425,"sourceType":"competition"},{"sourceId":9900,"sourceType":"datasetVersion","datasetId":6209}],"dockerImageVersionId":30558,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"
\n \n \n \n","metadata":{}},{"cell_type":"markdown","source":"#

Cats Vs Dogs Image Classification

\n","metadata":{}},{"cell_type":"markdown","source":"# Table of Content \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 1. Unzipping the train and test files\n
\n 2. Creating dataframes of image paths and labels\n
\n 3. Exploring The Data\n
\n 4. ImageDataGenerator\n
\n 5. ConvNN Sequential Model\n
\n 6. Transfer Learning: VGG16\n
\n\n","metadata":{}},{"cell_type":"markdown","source":"\n\n# Unzipping the train and test files ","metadata":{}},{"cell_type":"code","source":"FAST_RUN = False","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:14.218085Z","iopub.execute_input":"2024-05-22T03:05:14.218759Z","iopub.status.idle":"2024-05-22T03:05:14.227913Z","shell.execute_reply.started":"2024-05-22T03:05:14.218731Z","shell.execute_reply":"2024-05-22T03:05:14.227048Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import os\nfrom zipfile import ZipFile\n\n# printing all files-paths\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n for filename in filenames:\n print(os.path.join(dirname, filename))\n\npath = '/kaggle/input/dogs-vs-cats-redux-kernels-edition/'\nfiles = ['train.zip', 'test.zip']\n\n# unzipping the train and test files\nfor file in files:\n with ZipFile(path + file, 'r') as zip_file:\n zip_file.extractall(path = '/kaggle/working/')","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:26.847754Z","iopub.execute_input":"2024-05-22T03:05:26.848333Z","iopub.status.idle":"2024-05-22T03:05:41.738581Z","shell.execute_reply.started":"2024-05-22T03:05:26.848304Z","shell.execute_reply":"2024-05-22T03:05:41.737737Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n\n# Creating dataframes of image paths and labels \n\n**The Training dataframe consists of two columns :**\n- the file name for each image\n- the label for each iamge \"0 for cat, 1 for dogs\"","metadata":{}},{"cell_type":"code","source":"import pandas as pd\n\ntrain_path = '/kaggle/working/train/'\n\n# reading all filenames in train directory\nfilenames = os.listdir(train_path)\nprint(\"first filenames in the training images:\", filenames[:4], sep = \"\\n\")\n\n# extracting the label ('dog', 'cat') from filenames\nlabels = [int(x.split('.')[0] == 'dog') for x in filenames]\nprint(\"labels :\", labels[:4], sep = \"\\n\")\n\n# creating the train dataframe with filenames, labels\ntrain_df = pd.DataFrame({'filename' : filenames, 'label' : labels})\nprint('training set shape', train_df.shape, sep = \"\\n\")\n\ntrain_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:41.740114Z","iopub.execute_input":"2024-05-22T03:05:41.740395Z","iopub.status.idle":"2024-05-22T03:05:42.081378Z","shell.execute_reply.started":"2024-05-22T03:05:41.740369Z","shell.execute_reply":"2024-05-22T03:05:42.080515Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**The Testing dataframe with only one column `filename` and sorting it by id**","metadata":{}},{"cell_type":"code","source":"test_path = '/kaggle/working/test/'\n\n# reading all filenames in test directory\nfilenames = os.listdir(test_path)\nprint(\"first filenames in the test images:\", filenames[:4], sep = \"\\n\")\n\n# extracting the id from filename\nids = [int(x.split('.')[0]) for x in filenames]\nprint(\"ids :\", ids[:4], sep = \"\\n\")\n\n# creating the dataframe with ids, filenames\ntest = pd.DataFrame({'id' : ids, 'filename' : filenames})\nprint('test-set shape:', test.shape, sep = \"\\n\")\n\n# sorting the test dataframe by the id and reseting index\ntest.sort_values(by = ['id'], axis = 0, inplace = True)\ntest.reset_index(inplace = True, drop = True)\n\n# droping the id column as it's no longer useful\ntest.drop(columns = ['id'], inplace = True)\n\ntest.head()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:42.082562Z","iopub.execute_input":"2024-05-22T03:05:42.083031Z","iopub.status.idle":"2024-05-22T03:05:42.126023Z","shell.execute_reply.started":"2024-05-22T03:05:42.083003Z","shell.execute_reply":"2024-05-22T03:05:42.125185Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n\n# Explore the Data","metadata":{}},{"cell_type":"markdown","source":"**Spliting the data into train and validation**","metadata":{}},{"cell_type":"code","source":"from sklearn.model_selection import train_test_split\n\ntrain, validation = train_test_split(train_df, test_size = 0.2,\n stratify = train_df['label'], random_state = 42)\n\nprint(\"Train-set shape:\", train.shape)\nprint(\"Validation-set shape:\", validation.shape)","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:42.128254Z","iopub.execute_input":"2024-05-22T03:05:42.128866Z","iopub.status.idle":"2024-05-22T03:05:42.663641Z","shell.execute_reply.started":"2024-05-22T03:05:42.128831Z","shell.execute_reply":"2024-05-22T03:05:42.662700Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Importing visualization librarie and Setting up rcParams**","metadata":{}},{"cell_type":"code","source":"import matplotlib.pyplot as plt\nimport seaborn as sns\n%matplotlib inline\n\nplt.style.use('seaborn-v0_8-muted')\n\nplt.rcParams['font.size'] = 10\nplt.rcParams['legend.fontsize'] = 10\nplt.rcParams['xtick.labelsize'] = 8\nplt.rcParams['ytick.labelsize'] = 8","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:42.664814Z","iopub.execute_input":"2024-05-22T03:05:42.665165Z","iopub.status.idle":"2024-05-22T03:05:42.934101Z","shell.execute_reply.started":"2024-05-22T03:05:42.665126Z","shell.execute_reply":"2024-05-22T03:05:42.933124Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Cats and Dogs Classes are balanced**","metadata":{}},{"cell_type":"code","source":"fig, ax = plt.subplots(1, 2, figsize=(8, 3), sharey = True)\n\nsns.countplot(x = 'label', data = train, ax=ax[0])\nax[0].set_title('Training dataset')\nax[0].set_xticks(ticks=[0, 1], labels=['Cat', 'Dog'])\n\nsns.countplot(x = 'label', data = validation, ax=ax[1])\nax[1].set_title('Validation dataset')\nax[1].set_xticks(ticks=[0, 1], labels=['Cat', 'Dog'])\n\nplt.suptitle('Cats and Dogs Classes are balanced', y = -0.04, fontsize=18)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:42.935324Z","iopub.execute_input":"2024-05-22T03:05:42.935694Z","iopub.status.idle":"2024-05-22T03:05:43.294748Z","shell.execute_reply.started":"2024-05-22T03:05:42.935659Z","shell.execute_reply":"2024-05-22T03:05:43.293814Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**looking at a sample**","metadata":{}},{"cell_type":"code","source":"from tensorflow.keras.preprocessing.image import load_img\n\nload_img((train_path + train.loc[0, 'filename']))","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:05:57.198213Z","iopub.execute_input":"2024-05-22T03:05:57.199353Z","iopub.status.idle":"2024-05-22T03:06:04.855837Z","shell.execute_reply.started":"2024-05-22T03:05:57.199305Z","shell.execute_reply":"2024-05-22T03:06:04.854867Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n\n# ImageDataGenerator ","metadata":{}},{"cell_type":"markdown","source":"**defining ImageDataGenerator object to load images**","metadata":{}},{"cell_type":"code","source":"IMAGE_WIDTH = 360\nIMAGE_HEIGHT = 360\nIMAGE_SIZE = (IMAGE_WIDTH, IMAGE_HEIGHT)\nIMAGE_CHANNELS = 3\n\nBATCH_SIZE = 32","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:07.886595Z","iopub.execute_input":"2024-05-22T03:06:07.887710Z","iopub.status.idle":"2024-05-22T03:06:07.891943Z","shell.execute_reply.started":"2024-05-22T03:06:07.887675Z","shell.execute_reply":"2024-05-22T03:06:07.890906Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from tensorflow.keras.preprocessing.image import ImageDataGenerator\n\ntrain_datagen = ImageDataGenerator( rescale = 1.0/255, \n rotation_range = 15.0,\n width_shift_range = 0.1,\n height_shift_range = 0.1,\n shear_range = 0.1,\n zoom_range = 0.1,\n horizontal_flip= True,\n fill_mode = 'nearest',)\n\ntrain_datagen = train_datagen.flow_from_dataframe(\n # dataframe to get paths from\n dataframe = train,\n # directory path images are in\n directory = train_path,\n # column name file paths are in\n x_col = 'filename',\n # column name output targets are in\n y_col = 'label',\n # size to rescale the output images to\n target_size = IMAGE_SIZE,\n # size of the batch\n batch_size = BATCH_SIZE,\n # mode \"row\" output targets is array of y_col as it is\n class_mode = 'raw',\n )","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:13.754373Z","iopub.execute_input":"2024-05-22T03:06:13.755256Z","iopub.status.idle":"2024-05-22T03:06:13.964190Z","shell.execute_reply.started":"2024-05-22T03:06:13.755223Z","shell.execute_reply":"2024-05-22T03:06:13.963413Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Visualizing a sample from training images**","metadata":{}},{"cell_type":"code","source":"fig, ax = plt.subplots(4, 6, figsize = (12, 8))\n\ni, j = 0, 0\nfor images_batch, labels_batch in train_datagen:\n ax[i][j].imshow(images_batch[0])\n ax[i][j].set_title(['Cat', 'Dog'][labels_batch[0]])\n ax[i][j].axis('off')\n i += 1\n if i == 4: i, j = 0, j+1\n if j == 6: break\n \nfig.tight_layout(pad=0.001)\nplt.suptitle('Training Dataset', fontsize=18, y = 1.05)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:18.096385Z","iopub.execute_input":"2024-05-22T03:06:18.096776Z","iopub.status.idle":"2024-05-22T03:06:40.436656Z","shell.execute_reply.started":"2024-05-22T03:06:18.096743Z","shell.execute_reply":"2024-05-22T03:06:40.434680Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"validation_datagen = ImageDataGenerator(rescale = 1.0/255,)\n\nvalidation_datagen = validation_datagen.flow_from_dataframe(\n dataframe = validation,\n directory = train_path,\n x_col = 'filename',\n y_col = 'label',\n target_size = IMAGE_SIZE,\n batch_size = BATCH_SIZE,\n class_mode = 'raw',\n \n)","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:40.438287Z","iopub.execute_input":"2024-05-22T03:06:40.438606Z","iopub.status.idle":"2024-05-22T03:06:40.498308Z","shell.execute_reply.started":"2024-05-22T03:06:40.438576Z","shell.execute_reply":"2024-05-22T03:06:40.497469Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_datagen = ImageDataGenerator(rescale=1./255)\ntest_datagen = test_datagen.flow_from_dataframe(\n test, \n test_path, \n x_col='filename',\n y_col=None,\n class_mode=None,\n target_size= IMAGE_SIZE,\n batch_size= BATCH_SIZE,\n shuffle=False\n )","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:40.499617Z","iopub.execute_input":"2024-05-22T03:06:40.499977Z","iopub.status.idle":"2024-05-22T03:06:40.631189Z","shell.execute_reply.started":"2024-05-22T03:06:40.499941Z","shell.execute_reply":"2024-05-22T03:06:40.630447Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Visualizing a sample from testing images**","metadata":{}},{"cell_type":"code","source":"fig, ax = plt.subplots(4, 6, figsize = (13, 9))\nplt.subplots_adjust(hspace=0.05, wspace=0.05)\n\ni, j = 0, 0\nfor images_batch in test_datagen:\n ax[i][j].imshow(images_batch[0])\n ax[i][j].axis('off')\n i += 1\n if i == 4: i, j = 0, j+1\n if j == 6: break\n\nfig.tight_layout(pad=0.001)\nplt.suptitle('Testing Dataset', fontsize=18, y = 1.05)\n\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:40.633117Z","iopub.execute_input":"2024-05-22T03:06:40.633392Z","iopub.status.idle":"2024-05-22T03:06:44.758677Z","shell.execute_reply.started":"2024-05-22T03:06:40.633366Z","shell.execute_reply":"2024-05-22T03:06:44.757563Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n\n# ConvNN Sequential Model ","metadata":{}},{"cell_type":"markdown","source":"**building model structure**","metadata":{}},{"cell_type":"code","source":"!pip install visualkeras","metadata":{"_kg_hide-input":true,"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-05-22T03:06:44.760083Z","iopub.execute_input":"2024-05-22T03:06:44.760382Z","iopub.status.idle":"2024-05-22T03:06:57.642730Z","shell.execute_reply.started":"2024-05-22T03:06:44.760353Z","shell.execute_reply":"2024-05-22T03:06:57.641656Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import tensorflow as tf\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Conv2D, MaxPooling2D, Dropout, Flatten, Activation, BatchNormalization\nimport visualkeras\n\ntf.keras.backend.clear_session()\ntf.random.set_seed(42)\n\n\ndef build_model():\n model = Sequential()\n \n model.add(Conv2D(32, (3, 3), activation='relu',\n input_shape=(IMAGE_WIDTH, IMAGE_HEIGHT, IMAGE_CHANNELS, )))\n model.add(Conv2D(32, (3, 3), activation='relu'))\n model.add(BatchNormalization())\n model.add(MaxPooling2D(pool_size=(2, 2)))\n model.add(Dropout(0.5))\n\n model.add(Conv2D(64, (3, 3), activation='relu'))\n model.add(Conv2D(64, (3, 3), activation='relu'))\n model.add(BatchNormalization())\n model.add(MaxPooling2D(pool_size=(2, 2)))\n model.add(Dropout(0.5))\n\n model.add(Conv2D(128, (3, 3), activation='relu'))\n model.add(Conv2D(128, (3, 3), activation='relu'))\n model.add(BatchNormalization())\n model.add(MaxPooling2D(pool_size=(2, 2)))\n model.add(Dropout(0.5))\n\n model.add(Flatten())\n \n model.add(Dense(512, activation='relu'))\n model.add(BatchNormalization())\n model.add(Dropout(0.5))\n model.add(Dense(1, activation='sigmoid'))\n \n return model\n\nmodel = build_model()\nmodel.summary()\nvisualkeras.layered_view(model, legend = True, scale_xy=1, scale_z=0.001, spacing=15)","metadata":{"_kg_hide-input":false,"_kg_hide-output":false,"execution":{"iopub.status.busy":"2024-05-22T03:06:57.644594Z","iopub.execute_input":"2024-05-22T03:06:57.645012Z","iopub.status.idle":"2024-05-22T03:06:58.945337Z","shell.execute_reply.started":"2024-05-22T03:06:57.644961Z","shell.execute_reply":"2024-05-22T03:06:58.944433Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Compiling model**","metadata":{}},{"cell_type":"code","source":"adam_opt = tf.keras.optimizers.Adam(learning_rate=0.001)\nmodel.compile(optimizer = adam_opt, loss = 'binary_crossentropy', metrics = ['accuracy'])","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:58.946832Z","iopub.execute_input":"2024-05-22T03:06:58.947201Z","iopub.status.idle":"2024-05-22T03:06:58.965619Z","shell.execute_reply.started":"2024-05-22T03:06:58.947166Z","shell.execute_reply":"2024-05-22T03:06:58.964866Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Call-backs**","metadata":{}},{"cell_type":"code","source":"from keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint\n\n# Stop training when 'validation loss' has stopped improving.\nearly_stopping = EarlyStopping(monitor = 'val_loss', min_delta = 1e-4, \n patience = 5, verbose = 1)\n\n# Reduce learning rate when 'validation loss' has stopped improving.\nreduce_learning_rate = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.5,\n patience = 2, verbose = 1, min_lr = 1e-5)\n\n\n#saving best model weights on the validation-set\ncheckpoint_filepath = '/kaggle/working/checkpoint.weights.h5'\nmodel_checkpoint = ModelCheckpoint(\n filepath=checkpoint_filepath,\n save_weights_only=True,\n monitor='val_loss',\n mode='max',\n save_best_only=True\n )\n\n\n\ncall_backs = [reduce_learning_rate, early_stopping, model_checkpoint]","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:06:58.966839Z","iopub.execute_input":"2024-05-22T03:06:58.967185Z","iopub.status.idle":"2024-05-22T03:06:58.974033Z","shell.execute_reply.started":"2024-05-22T03:06:58.967151Z","shell.execute_reply":"2024-05-22T03:06:58.972923Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"epochs = 3 if FAST_RUN else 20\nhistory = model.fit(train_datagen,\n steps_per_epoch= train.shape[0]//BATCH_SIZE,\n epochs = epochs,\n validation_data= validation_datagen, \n validation_steps= validation.shape[0]// BATCH_SIZE,\n callbacks = call_backs,\n )","metadata":{"_kg_hide-output":true,"execution":{"iopub.status.busy":"2024-05-21T16:49:14.820806Z","iopub.execute_input":"2024-05-21T16:49:14.821122Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def plot_history(history):\n acc = history.history['accuracy']\n val_acc = history.history['val_accuracy']\n\n loss = history.history['loss']\n val_loss = history.history['val_loss']\n \n fig, ax = plt.subplots(2, 1, figsize = (8, 10))\n\n epochs = range(1, len(acc) + 1)\n ax[0].plot(epochs, acc, '--', label='Training acc')\n ax[0].plot(epochs, val_acc, label='Validation acc')\n ax[0].set_title('Training and validation accuracy')\n ax[0].set_ylim([0.0, 1.0])\n ax[0].legend()\n\n ax[1].plot(epochs, loss, '--', label='Training loss')\n ax[1].plot(epochs, val_loss, label='Validation loss')\n ax[1].set_title('Training and validation loss')\n ax[1].legend()\n \n plt.show()\n","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2024-05-22T03:07:34.287023Z","iopub.execute_input":"2024-05-22T03:07:34.287982Z","iopub.status.idle":"2024-05-22T03:07:34.296359Z","shell.execute_reply.started":"2024-05-22T03:07:34.287941Z","shell.execute_reply":"2024-05-22T03:07:34.295233Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Model Learning Curve**","metadata":{}},{"cell_type":"code","source":"plot_history(history)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Making Predictions**","metadata":{}},{"cell_type":"code","source":"import numpy as np\n\nmodel.load_weights(checkpoint_filepath)\npredict = model.predict(test_datagen, steps= np.ceil(test.shape[0]/BATCH_SIZE))\npredict","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sub = test.copy()\n\nsub['label'] = predict\nsub['id'] = sub['filename'].str.split('.').apply(lambda x: x[0])\nsub = sub[['id', 'label']]\nsub.to_csv('/kaggle/working/sub0.csv', index = False)\n\nsub.head()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"\n\n# Transfer Learning: VGG16 ","metadata":{}},{"cell_type":"code","source":"from keras.applications.vgg16 import VGG16\n\nvgg16=VGG16(\n include_top=False,\n pooling='max',\n input_shape =(IMAGE_WIDTH, IMAGE_HEIGHT, IMAGE_CHANNELS), \n weights='imagenet',\n)\nvgg16.trainable = False\n\nmodel = Sequential()\nmodel.add(vgg16)\nmodel.add(Dropout(0.5))\nmodel.add(Flatten())\nmodel.add(Dense(512, activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(Dropout(0.5))\nmodel.add(Dense(1, activation='sigmoid'))\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:08:39.004261Z","iopub.execute_input":"2024-05-22T03:08:39.005198Z","iopub.status.idle":"2024-05-22T03:08:39.476869Z","shell.execute_reply.started":"2024-05-22T03:08:39.005161Z","shell.execute_reply":"2024-05-22T03:08:39.475800Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Compile model**","metadata":{}},{"cell_type":"code","source":"adam_opt = tf.keras.optimizers.Adam(learning_rate=0.001)\nmodel.compile(optimizer = adam_opt, loss = 'binary_crossentropy', metrics = ['accuracy'])","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:07:56.638676Z","iopub.execute_input":"2024-05-22T03:07:56.639368Z","iopub.status.idle":"2024-05-22T03:07:56.651589Z","shell.execute_reply.started":"2024-05-22T03:07:56.639333Z","shell.execute_reply":"2024-05-22T03:07:56.650647Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Stop training when 'validation loss' has stopped improving.\nearly_stopping = EarlyStopping(monitor = 'val_loss', min_delta = 1e-4, \n patience = 5, verbose = 1)\n\n# Reduce learning rate when 'validation loss' has stopped improving.\nreduce_learning_rate = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.5,\n patience = 2, verbose = 1, min_lr = 1e-5)\n\n#saving best model weights on the validation-set\ncheckpoint_filepath = '/kaggle/working/checkpoint.weights.h5'\nmodel_checkpoint = ModelCheckpoint(\n filepath=checkpoint_filepath,\n save_weights_only=True,\n monitor='val_loss',\n mode='max',\n save_best_only=True\n )\n\n\n\ncall_backs = [reduce_learning_rate, early_stopping, model_checkpoint]","metadata":{"execution":{"iopub.status.busy":"2024-05-22T03:09:49.858913Z","iopub.execute_input":"2024-05-22T03:09:49.859297Z","iopub.status.idle":"2024-05-22T03:09:49.865939Z","shell.execute_reply.started":"2024-05-22T03:09:49.859266Z","shell.execute_reply":"2024-05-22T03:09:49.864874Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"**Training Model**","metadata":{}},{"cell_type":"code","source":"epochs = 3 if FAST_RUN else 20\nhistory = model.fit(train_datagen,\n steps_per_epoch= train.shape[0]//BATCH_SIZE,\n epochs = epochs,\n validation_data= validation_datagen, \n validation_steps= validation.shape[0]// BATCH_SIZE,\n callbacks = call_backs\n )","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plot_history(history)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import numpy as np\n\nmodel.load_weights(checkpoint_filepath)\npredict = model.predict(test_datagen, steps= np.ceil(test.shape[0]/BATCH_SIZE))\n\nsub = test.copy()\n\nsub['label'] = predict\nsub['id'] = sub['filename'].str.split('.').apply(lambda x: x[0])\nsub = sub[['id', 'label']]\nsub.to_csv('/kaggle/working/sub1.csv', index = False)\n\nsub.head()","metadata":{"trusted":true},"execution_count":null,"outputs":[]}]}