{"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":"code","source":"#imported the neccesary library's\nimport numpy as np\nimport pandas as pd\nimport tensorflow as tf\nimport matplotlib.pyplot as plt\nimport matplotlib.cm as cm\nfrom matplotlib.colors import Normalize\nimport seaborn as sns \nimport os \nfrom skimage import io\nfrom skimage.color import rgb2gray\nimport plotly.express as px\nimport cv2\nimport os\n\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.utils import to_categorical\nfrom tensorflow.keras import Sequential,activations, Input, Model, models\nfrom tensorflow.keras.callbacks import ReduceLROnPlateau\nfrom tensorflow.keras.layers import Dense, GlobalAveragePooling2D\nfrom tensorflow.keras.layers import AveragePooling2D, MaxPooling2D, Dropout\nfrom tensorflow.keras.layers import LeakyReLU\nfrom distutils.dir_util import copy_tree\n\nimport warnings\nwarnings.simplefilter('ignore')\nfrom tensorflow.keras.layers import Dense,Activation, Dropout, Conv2D, MaxPool2D, Flatten,GlobalAveragePooling2D, GlobalMaxPooling2D, BatchNormalization","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2022-05-02T14:51:28.278230Z","iopub.execute_input":"2022-05-02T14:51:28.278531Z","iopub.status.idle":"2022-05-02T14:51:31.757848Z","shell.execute_reply.started":"2022-05-02T14:51:28.278448Z","shell.execute_reply":"2022-05-02T14:51:31.757090Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#read the train csv file and display first 5 rows of csv file\ntrain=pd.read_csv('../input/happy-whale-and-dolphin/train.csv', header='infer')\ntrain.head()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:31.759075Z","iopub.execute_input":"2022-05-02T14:51:31.759311Z","iopub.status.idle":"2022-05-02T14:51:31.828681Z","shell.execute_reply.started":"2022-05-02T14:51:31.759278Z","shell.execute_reply":"2022-05-02T14:51:31.827879Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#display information about train csv file\ntrain.info()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:31.830196Z","iopub.execute_input":"2022-05-02T14:51:31.830454Z","iopub.status.idle":"2022-05-02T14:51:31.859546Z","shell.execute_reply.started":"2022-05-02T14:51:31.830419Z","shell.execute_reply":"2022-05-02T14:51:31.858806Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#display the no of missing values\nprint('Number of Missing Data:')\ntrain.isna().sum()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:31.862038Z","iopub.execute_input":"2022-05-02T14:51:31.862412Z","iopub.status.idle":"2022-05-02T14:51:31.885482Z","shell.execute_reply.started":"2022-05-02T14:51:31.862369Z","shell.execute_reply":"2022-05-02T14:51:31.884765Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#describe the species and individual_id\ntrain[['species', 'individual_id']].describe()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:31.886979Z","iopub.execute_input":"2022-05-02T14:51:31.887411Z","iopub.status.idle":"2022-05-02T14:51:31.927897Z","shell.execute_reply.started":"2022-05-02T14:51:31.887369Z","shell.execute_reply":"2022-05-02T14:51:31.927082Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(\"Total species before finding duplicates :\",len(train.species.unique()))\ntrain.species = train.species.str.replace('kiler_whale','killer_whale')\ntrain.species = train.species.str.replace('bottlenose_dolpin','bottlenose_dolphin')\ntrain['species'][(train['species'] ==\"pilot_whale\") | (train['species'] ==\"globis\" )]='short_finned_pilot_whale'\nprint(\"Total species after :\",len(train.species.unique()))","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:31.929345Z","iopub.execute_input":"2022-05-02T14:51:31.929607Z","iopub.status.idle":"2022-05-02T14:51:32.032241Z","shell.execute_reply.started":"2022-05-02T14:51:31.929573Z","shell.execute_reply":"2022-05-02T14:51:32.031407Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#display the Occurences of different species\nanimal_cnt = train.species.value_counts()\nprint(\"Occurences of different species:\")\nprint(animal_cnt)\nprint(f\"Total number of species: {len(animal_cnt)}\")","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.033982Z","iopub.execute_input":"2022-05-02T14:51:32.034275Z","iopub.status.idle":"2022-05-02T14:51:32.046471Z","shell.execute_reply.started":"2022-05-02T14:51:32.034237Z","shell.execute_reply":"2022-05-02T14:51:32.045778Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#add new column in train dataframe that indiacate the specie is whale or dolphin\ntrain['label'] = train.species.map(lambda x: 'dolphin' if 'dolphin' in x else 'whale')\ntrain.head()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.047857Z","iopub.execute_input":"2022-05-02T14:51:32.048336Z","iopub.status.idle":"2022-05-02T14:51:32.073524Z","shell.execute_reply.started":"2022-05-02T14:51:32.048298Z","shell.execute_reply":"2022-05-02T14:51:32.072859Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#display the no of whale's and dolphone's\ntrain['label'].value_counts()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.074803Z","iopub.execute_input":"2022-05-02T14:51:32.075272Z","iopub.status.idle":"2022-05-02T14:51:32.087927Z","shell.execute_reply.started":"2022-05-02T14:51:32.075236Z","shell.execute_reply":"2022-05-02T14:51:32.087177Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#plot the bar graph\nplt.bar(train['label'].value_counts().index, train['label'].value_counts().values)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.089209Z","iopub.execute_input":"2022-05-02T14:51:32.089644Z","iopub.status.idle":"2022-05-02T14:51:32.258941Z","shell.execute_reply.started":"2022-05-02T14:51:32.089608Z","shell.execute_reply":"2022-05-02T14:51:32.258108Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#plot the pie chart\nplt.pie(train['label'].value_counts().values,labels=train['label'].unique(),startangle = 90,autopct='%1.1f%%')\nplt.legend()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.263339Z","iopub.execute_input":"2022-05-02T14:51:32.266194Z","iopub.status.idle":"2022-05-02T14:51:32.428814Z","shell.execute_reply.started":"2022-05-02T14:51:32.266147Z","shell.execute_reply":"2022-05-02T14:51:32.427899Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#plot the horizontal bar graph of the speecies\nplt.figure(figsize=(10, 10))\nplt.barh(train['species'].value_counts().index.str.replace('_', ' '), train['species'].value_counts().values, color=(0.2, 0.4, 0.6, 0.6))","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.433459Z","iopub.execute_input":"2022-05-02T14:51:32.433763Z","iopub.status.idle":"2022-05-02T14:51:32.850224Z","shell.execute_reply.started":"2022-05-02T14:51:32.433720Z","shell.execute_reply":"2022-05-02T14:51:32.849502Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_images_path = \"../input/jpeg-happywhale-128x128/train_images-128-128/train_images-128-128\"","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.854243Z","iopub.execute_input":"2022-05-02T14:51:32.854860Z","iopub.status.idle":"2022-05-02T14:51:32.860768Z","shell.execute_reply.started":"2022-05-02T14:51:32.854816Z","shell.execute_reply":"2022-05-02T14:51:32.859946Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#label contains the individual_id\nlabel=train[\"individual_id\"].unique().tolist()\n#label.append(\"new_individual\")\nlabel=np.array(label)\nlabel.shape","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.905924Z","iopub.execute_input":"2022-05-02T14:51:32.906399Z","iopub.status.idle":"2022-05-02T14:51:32.926632Z","shell.execute_reply.started":"2022-05-02T14:51:32.906362Z","shell.execute_reply":"2022-05-02T14:51:32.925884Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#copy input train directory to the temp directory\nt=copy_tree(\"../input/jpeg-happywhale-128x128/train_images-128-128/train_images-128-128\",\"/kaggle/temp/train/\")","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:51:32.928902Z","iopub.execute_input":"2022-05-02T14:51:32.929383Z","iopub.status.idle":"2022-05-02T14:52:24.038099Z","shell.execute_reply.started":"2022-05-02T14:51:32.929342Z","shell.execute_reply":"2022-05-02T14:52:24.037283Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#make validation directory\nif not os.path.exists(\"/kaggle/temp/valid/\"):\n os.mkdir(\"/kaggle/temp/valid/\")","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:24.052508Z","iopub.execute_input":"2022-05-02T14:52:24.052751Z","iopub.status.idle":"2022-05-02T14:52:24.073510Z","shell.execute_reply.started":"2022-05-02T14:52:24.052716Z","shell.execute_reply":"2022-05-02T14:52:24.072482Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#make the directory of the each individual_id in train and validation\nfor i in label:\n if not os.path.exists(\"/kaggle/temp/train/\"+i):\n os.mkdir(\"/kaggle/temp/train/\"+i)\n if not os.path.exists(\"/kaggle/temp/valid/\"+i):\n os.mkdir(\"/kaggle/temp/valid/\"+i)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:24.075055Z","iopub.execute_input":"2022-05-02T14:52:24.075803Z","iopub.status.idle":"2022-05-02T14:52:24.207867Z","shell.execute_reply.started":"2022-05-02T14:52:24.075763Z","shell.execute_reply":"2022-05-02T14:52:24.207071Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#split train data and validation data in 50 % and 50% \nfor i in range(len(train[\"image\"])):\n img=train[\"image\"][i]\n t= np.random.random()\n \n if t<= 0.5:\n new_path= os.path.join(\"/kaggle/temp/train/\",train[\"individual_id\"][i],img)\n os.replace('/kaggle/temp/train/'+img,new_path)\n else:\n new_path= os.path.join(\"/kaggle/temp/valid/\",train[\"individual_id\"][i],img)\n os.replace('/kaggle/temp/train/'+img,new_path)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:24.209072Z","iopub.execute_input":"2022-05-02T14:52:24.209687Z","iopub.status.idle":"2022-05-02T14:52:27.576947Z","shell.execute_reply.started":"2022-05-02T14:52:24.209651Z","shell.execute_reply":"2022-05-02T14:52:27.576114Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#apply the image generator on train images for augmention\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=20,\n zoom_range=0.4,\n width_shift_range=0.4,\n height_shift_range=0.4,\n horizontal_flip=False,\n vertical_flip=False,\n rescale=1./255\n )","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:27.578241Z","iopub.execute_input":"2022-05-02T14:52:27.578487Z","iopub.status.idle":"2022-05-02T14:52:27.587300Z","shell.execute_reply.started":"2022-05-02T14:52:27.578455Z","shell.execute_reply":"2022-05-02T14:52:27.586708Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_gen = train_datagen.flow_from_directory(directory=\"/kaggle/temp/train/\", target_size=(128,128),batch_size=64)\nvalidation_datagen = ImageDataGenerator(rescale=1./255)\nvalid_gen = validation_datagen.flow_from_directory(directory='/kaggle/temp/valid/',target_size=(128,128),batch_size=64)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:27.590766Z","iopub.execute_input":"2022-05-02T14:52:27.593016Z","iopub.status.idle":"2022-05-02T14:52:34.449992Z","shell.execute_reply.started":"2022-05-02T14:52:27.592971Z","shell.execute_reply":"2022-05-02T14:52:34.449186Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#build the model\n\n\nmodel=Sequential()\n#Convolution layer 1\nmodel.add(Conv2D(filters=32, kernel_size=(5,5), strides=(1,1), padding='valid',input_shape=(128,128, 3)))\nmodel.add(BatchNormalization())\nmodel.add(Activation(LeakyReLU(alpha=0.01)))\n\nmodel.add(AveragePooling2D((3, 3)))\nmodel.add(Conv2D(filters=64, kernel_size=(4,4), strides=(1,1), padding='valid'))\nmodel.add(BatchNormalization())\nmodel.add(Activation(LeakyReLU(alpha=0.01)))\nmodel.add(AveragePooling2D((3, 3)))\n#Convolution layer 2\nmodel.add(Conv2D(filters=96, kernel_size=(3,3), strides=(1,1), padding='valid'))\nmodel.add(BatchNormalization())\nmodel.add(Activation(LeakyReLU(alpha=0.01)))\nmodel.add(AveragePooling2D((3, 3)))\n\n\nmodel.add(Conv2D(filters=128, kernel_size=(2,2), strides=(1,1), padding='valid'))\nmodel.add(BatchNormalization())\nmodel.add(Activation(LeakyReLU(alpha=0.01)))\n\nmodel.add(Conv2D(filters=15587, kernel_size=(2,2), strides=(1,1), padding='valid'))\nmodel.add(Activation('softmax'))\nmodel.add(Flatten())\n\nmodel.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:34.459624Z","iopub.execute_input":"2022-05-02T14:52:34.460104Z","iopub.status.idle":"2022-05-02T14:52:35.731720Z","shell.execute_reply.started":"2022-05-02T14:52:34.460052Z","shell.execute_reply":"2022-05-02T14:52:35.731014Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#setup callback\nreduceLROnPlateau = ReduceLROnPlateau(monitor='val_acc',patience=5,verbose=1,factor=0.6,min_lr=0.0001)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:35.732792Z","iopub.execute_input":"2022-05-02T14:52:35.733232Z","iopub.status.idle":"2022-05-02T14:52:35.737329Z","shell.execute_reply.started":"2022-05-02T14:52:35.733192Z","shell.execute_reply":"2022-05-02T14:52:35.736603Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#train the model \nhistory=model.fit(train_gen,epochs=200,callbacks=[reduceLROnPlateau],validation_data=valid_gen,steps_per_epoch=train_gen.n//train_gen.batch_size,\n validation_steps= valid_gen.n//valid_gen.batch_size,workers=8,use_multiprocessing=True)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T14:52:35.738707Z","iopub.execute_input":"2022-05-02T14:52:35.738996Z","iopub.status.idle":"2022-05-02T15:25:23.684162Z","shell.execute_reply.started":"2022-05-02T14:52:35.738957Z","shell.execute_reply":"2022-05-02T15:25:23.683048Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# display the Accuracy vs Epoch plot\nplt.figure(figsize=(15,5))\nplt.plot(history.history['accuracy'],label='train')\nplt.plot(history.history['val_accuracy'],label='valid')\n\nplt.title('Report of Accuracy')\nplt.ylabel('Accuracy')\nplt.xlabel('Epoch')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:32:18.597607Z","iopub.execute_input":"2022-05-02T15:32:18.597896Z","iopub.status.idle":"2022-05-02T15:32:18.790327Z","shell.execute_reply.started":"2022-05-02T15:32:18.597862Z","shell.execute_reply":"2022-05-02T15:32:18.789623Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# display the Loss vs Epoch plot\nplt.figure(figsize=(15,5))\nplt.plot(history.history['loss'],label='train')\nplt.plot(history.history['val_loss'],label='validation')\n\nplt.title('Report of Loss')\nplt.ylabel('loss')\nplt.xlabel('Epoch')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:33:01.165278Z","iopub.execute_input":"2022-05-02T15:33:01.166193Z","iopub.status.idle":"2022-05-02T15:33:01.343331Z","shell.execute_reply.started":"2022-05-02T15:33:01.166153Z","shell.execute_reply":"2022-05-02T15:33:01.342676Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# test folder copy into the temp directory\np=copy_tree(\"../input/jpeg-happywhale-128x128/test_images-128-128/test_images-128-128\",\"/kaggle/temp/test/test_img/\")","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:25:24.372684Z","iopub.execute_input":"2022-05-02T15:25:24.372933Z","iopub.status.idle":"2022-05-02T15:28:13.601642Z","shell.execute_reply.started":"2022-05-02T15:25:24.372899Z","shell.execute_reply":"2022-05-02T15:28:13.600888Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#apply image genetory on test images\ntest_datagen = ImageDataGenerator(rescale=1./255)\ntest_gen = test_datagen.flow_from_directory(directory='/kaggle/temp/test/',target_size=(128,128),batch_size=64\n ,class_mode=None,shuffle=False)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:28:13.617783Z","iopub.execute_input":"2022-05-02T15:28:13.618123Z","iopub.status.idle":"2022-05-02T15:28:14.310624Z","shell.execute_reply.started":"2022-05-02T15:28:13.618079Z","shell.execute_reply":"2022-05-02T15:28:14.309905Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#prediction\npredictions= model.predict(test_gen)\npredictions.shape","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:28:14.311932Z","iopub.execute_input":"2022-05-02T15:28:14.312334Z","iopub.status.idle":"2022-05-02T15:28:55.552232Z","shell.execute_reply.started":"2022-05-02T15:28:14.312297Z","shell.execute_reply":"2022-05-02T15:28:55.551559Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#fetch and store the highest five probability values\nprediction_prob=[]\n\nfor i in range(len(predictions)):\n sort_prob=np.sort(predictions[i])[::-1][:5]\n prediction_prob.append(sort_prob)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:28:55.553745Z","iopub.execute_input":"2022-05-02T15:28:55.554020Z","iopub.status.idle":"2022-05-02T15:29:24.916403Z","shell.execute_reply.started":"2022-05-02T15:28:55.553986Z","shell.execute_reply":"2022-05-02T15:29:24.915637Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#fetch and store the index of the high five probability values\nprediction_index=[]\nfor i in range(len(predictions)):\n sort_index=np.argsort(predictions[i])[::-1][:5]\n prediction_index.append(sort_index)","metadata":{"execution":{"iopub.status.busy":"2022-05-02T15:29:24.917648Z","iopub.execute_input":"2022-05-02T15:29:24.917980Z","iopub.status.idle":"2022-05-02T15:29:58.451789Z","shell.execute_reply.started":"2022-05-02T15:29:24.917943Z","shell.execute_reply":"2022-05-02T15:29:58.451030Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#limit is threshhold value of new_individual id\nlimit=0.7\n\n#compare with all prob value to the limit if less than swap \nprediction_lite=[]\nfor i in range(len(prediction_index)):\n if prediction_prob[i][0]