{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:51:02.586061Z", "iopub.status.busy": "2023-05-31T11:51:02.585717Z", "iopub.status.idle": "2023-05-31T11:51:02.593440Z", "shell.execute_reply": "2023-05-31T11:51:02.592394Z", "shell.execute_reply.started": "2023-05-31T11:51:02.586032Z" } }, "outputs": [], "source": [ "import numpy as np\n", "import pandas as pd\n", "import matplotlib.pyplot as plt\n", "\n", "from tqdm import tqdm\n", "from tensorflow.keras.preprocessing import image\n", "from sklearn.preprocessing import label_binarize\n", "from sklearn.model_selection import train_test_split\n", "from keras.models import Sequential\n", "from keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D, Dropout\n", "from keras.optimizers import Adam" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.667786Z", "iopub.status.busy": "2023-05-31T11:55:45.667374Z", "iopub.status.idle": "2023-05-31T11:55:45.822595Z", "shell.execute_reply": "2023-05-31T11:55:45.821479Z", "shell.execute_reply.started": "2023-05-31T11:55:45.667757Z" } }, "outputs": [], "source": [ "import tensorflow as tf\n", "from keras.applications.resnet_v2 import ResNet50V2\n", "from tensorflow import keras\n", "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n", "import seaborn as sns\n", "import matplotlib.pyplot as plt\n", "import cv2\n" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.825161Z", "iopub.status.busy": "2023-05-31T11:55:45.824796Z", "iopub.status.idle": "2023-05-31T11:55:45.829968Z", "shell.execute_reply": "2023-05-31T11:55:45.828863Z", "shell.execute_reply.started": "2023-05-31T11:55:45.825128Z" } }, "outputs": [], "source": [ "import os\n", "from tensorflow.keras.preprocessing.image import load_img" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.831997Z", "iopub.status.busy": "2023-05-31T11:55:45.831637Z", "iopub.status.idle": "2023-05-31T11:55:45.860368Z", "shell.execute_reply": "2023-05-31T11:55:45.859192Z", "shell.execute_reply.started": "2023-05-31T11:55:45.831961Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "(10222, 2)\n" ] }, { "data": { "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", "
idbreed
0000bec180eb18c7604dcecc8fe0dba07boston_bull
1001513dfcb2ffafc82cccf4d8bbaba97dingo
2001cdf01b096e06d78e9e5112d419397pekinese
300214f311d5d2247d5dfe4fe24b2303dbluetick
40021f9ceb3235effd7fcde7f7538ed62golden_retriever
\n", "
" ], "text/plain": [ " id breed\n", "0 000bec180eb18c7604dcecc8fe0dba07 boston_bull\n", "1 001513dfcb2ffafc82cccf4d8bbaba97 dingo\n", "2 001cdf01b096e06d78e9e5112d419397 pekinese\n", "3 00214f311d5d2247d5dfe4fe24b2303d bluetick\n", "4 0021f9ceb3235effd7fcde7f7538ed62 golden_retriever" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "labels_all = pd.read_csv('data_small/New folder/labels.csv')\n", "print(labels_all.shape)\n", "labels_all.head()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.863383Z", "iopub.status.busy": "2023-05-31T11:55:45.862940Z", "iopub.status.idle": "2023-05-31T11:55:45.886516Z", "shell.execute_reply": "2023-05-31T11:55:45.885072Z", "shell.execute_reply.started": "2023-05-31T11:55:45.863351Z" } }, "outputs": [], "source": [ "breed_all = labels_all['breed']\n", "breed_count = breed_all.value_counts()\n", "breed_count.head()" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.890591Z", "iopub.status.busy": "2023-05-31T11:55:45.888452Z", "iopub.status.idle": "2023-05-31T11:55:45.924310Z", "shell.execute_reply": "2023-05-31T11:55:45.916796Z", "shell.execute_reply.started": "2023-05-31T11:55:45.890548Z" } }, "outputs": [ { "data": { "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", " \n", " \n", " \n", " \n", " \n", " \n", "
indexidbreed
090042188c895a2f14ef64a918ed9c7b64scottish_deerhound
11200693b8bc2470375cc744a6391d397ecmaltese_dog
27901e787576c003930f96c966f9c3e1d44scottish_deerhound
38001ee3c7ff9bcaba9874183135877670eentlebucher
488021b5a49189665c0442c19b5b33e8cf1entlebucher
\n", "
" ], "text/plain": [ " index id breed\n", "0 9 0042188c895a2f14ef64a918ed9c7b64 scottish_deerhound\n", "1 12 00693b8bc2470375cc744a6391d397ec maltese_dog\n", "2 79 01e787576c003930f96c966f9c3e1d44 scottish_deerhound\n", "3 80 01ee3c7ff9bcaba9874183135877670e entlebucher\n", "4 88 021b5a49189665c0442c19b5b33e8cf1 entlebucher" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "CLASS_NAME = ['scottish_deerhound', 'maltese_dog', 'afghan_hound', 'entlebucher', 'bernese_mountain_dog']\n", "labels = labels_all[(labels_all['breed'].isin(CLASS_NAME))]\n", "labels = labels.reset_index()\n", "labels.head()" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:45.934744Z", "iopub.status.busy": "2023-05-31T11:55:45.931530Z", "iopub.status.idle": "2023-05-31T11:55:45.955887Z", "shell.execute_reply": "2023-05-31T11:55:45.954669Z", "shell.execute_reply.started": "2023-05-31T11:55:45.934705Z" } }, "outputs": [], "source": [ "train_path = 'data_small/New folder/train'\n", "\n", "\n", "#reading dataset labels\n", "train_labels = pd.read_csv('data_small/New folder/labels.csv')\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:46.081478Z", "iopub.status.busy": "2023-05-31T11:55:46.080604Z", "iopub.status.idle": "2023-05-31T11:55:46.105439Z", "shell.execute_reply": "2023-05-31T11:55:46.103383Z", "shell.execute_reply.started": "2023-05-31T11:55:46.081439Z" } }, "outputs": [], "source": [ "train_labels.head()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T12:01:44.672309Z", "iopub.status.busy": "2023-05-31T12:01:44.671926Z", "iopub.status.idle": "2023-05-31T12:01:44.723863Z", "shell.execute_reply": "2023-05-31T12:01:44.722593Z", "shell.execute_reply.started": "2023-05-31T12:01:44.672280Z" } }, "outputs": [], "source": [ "# Create X_data as a pandas dataframe\n", "X_data = pd.DataFrame(columns=['image'])\n", "\n", "# Loop over the images and load them into X_data\n", "for i, file_path in enumerate(train_labels):\n", " img = cv2.imread(file_path)\n", " print(i, file_path, img)\n", " x = np.expand_dims(img,axis=0)\n", " print(i, x)\n", " X_data.loc[i] = x[i] / 255.0\n", "\n", "# Create train_labels as a pandas dataframe\n", "train_labels = pd.read_csv('/kaggle/input/dog-breeding/New folder/train')\n", "\n", "# Merge X_data and train_labels on the 'id' column\n", "train_data = pd.merge(X_data, train_labels, on='id')\n", "\n", "# Print train image and one hot encode shape & size\n", "print('\\nTrain Images shape:', X_data.shape, ' size: {:,}'.format(X_data.size))" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:55:48.195541Z", "iopub.status.busy": "2023-05-31T11:55:48.194955Z", "iopub.status.idle": "2023-05-31T11:55:48.400186Z", "shell.execute_reply": "2023-05-31T11:55:48.399006Z", "shell.execute_reply.started": "2023-05-31T11:55:48.195507Z" } }, "outputs": [], "source": [ "x = train_labels.breed.unique()\n", "include = ['beagle', 'chihuahua', 'doberman','french_bulldog', 'golden_retriever', 'malamute', 'pug', 'saint_bernard', 'scottish_deerhound','tibetan_mastiff']\n", "for i in x :\n", " if i not in include:\n", " train_labels = train_labels.drop(train_labels[ train_labels['breed'] == i ].index)\n", "print(train_labels.shape)\n", "train_labels.head()" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "execution": { "iopub.execute_input": "2023-05-31T11:51:03.032746Z", "iopub.status.busy": "2023-05-31T11:51:03.029528Z", "iopub.status.idle": "2023-05-31T11:51:04.762119Z", "shell.execute_reply": "2023-05-31T11:51:04.759538Z", "shell.execute_reply.started": "2023-05-31T11:51:03.032713Z" } }, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "100%|██████████| 588/588 [00:00<00:00, 630.73it/s]" ] }, { "name": "stdout", "output_type": "stream", "text": [ "\n", "Train Images shape: (3, 1, 224, 224, 3) size: 451,584\n", "One-hot encoded output shape: (3, 5) size: 15\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "\n" ] } ], "source": [ "# fix --- different data structure to store imgs and ids\n", "# X_data = np.zeros((len(labels), 224, 224, 3), dtype='float32')\n", "X_data = []\n", "ids = []\n", "valid_indices = []\n", "\n", "# One hot encoding\n", "Y_data = label_binarize(labels['breed'], classes = CLASS_NAME)\n", "\n", "# Reading and converting image to numpy array and normalizing dataset\n", "for i in tqdm(range(len(labels))):\n", " try: # for fast reproducing and fixing purposes (because of sampled data)\n", " img = image.load_img(f'data_small/New folder/train/{labels[\"id\"][i]}.jpg', target_size=(224, 224))\n", " except FileNotFoundError:\n", " continue\n", " img = image.img_to_array(img)\n", " \n", " \n", " x = np.expand_dims(img.copy(), axis=0)\n", " \n", " # fix --- different data structure to store imgs and ids\n", " X_data.append(x / 255.0)\n", " ids.append(labels.iloc[i]['id'])\n", " valid_indices.append(i)\n", "\n", "X_data = np.array(X_data)\n", "Y_data = Y_data[valid_indices]\n", "\n", "# Printing train image and one hot encode shape & size\n", "print('\\nTrain Images shape: ',X_data.shape,' size: {:,}'.format(X_data.size))\n", "print('One-hot encoded output shape: ',Y_data.shape,' size: {:,}'.format(Y_data.size))" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.763647Z", "iopub.status.idle": "2023-05-31T11:51:04.764242Z", "shell.execute_reply": "2023-05-31T11:51:04.763903Z", "shell.execute_reply.started": "2023-05-31T11:51:04.763875Z" } }, "outputs": [], "source": [ "plt.figure(figsize=(18,4))\n", "cp = sns.countplot(x = 'breed', data = train_labels)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.766193Z", "iopub.status.idle": "2023-05-31T11:51:04.766701Z", "shell.execute_reply": "2023-05-31T11:51:04.766459Z", "shell.execute_reply.started": "2023-05-31T11:51:04.766438Z" } }, "outputs": [], "source": [ "invalid_files = []\n", "for filename in filenames:\n", " try:\n", " load_img(os.path.join(data_dir, filename))\n", " except:\n", " invalid_files.append(filename)\n", "\n", "print(f\"Number of invalid files: {len(invalid_files)}\")\n", "print(\"Invalid files: \", invalid_files)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.769183Z", "iopub.status.idle": "2023-05-31T11:51:04.769966Z", "shell.execute_reply": "2023-05-31T11:51:04.769748Z", "shell.execute_reply.started": "2023-05-31T11:51:04.769725Z" } }, "outputs": [], "source": [ "train_datagen = ImageDataGenerator(\n", " rescale = 1./255, \n", " validation_split = 0.2,\n", " shear_range = 0.2, \n", " zoom_range = 0.2, horizontal_flip = True, \n", " width_shift_range = 0.2,\n", " height_shift_range = 0.2,\n", " rotation_range = 20,\n", " brightness_range=[0.2,1.0])\n", "\n", "train_set = train_datagen.flow_from_dataframe(dataframe = train_labels,\n", " directory = train_path,\n", " x_col = \"id\",\n", " y_col = \"breed\",\n", " batch_size = 16,\n", " subset=\"training\",\n", " class_mode = \"categorical\",\n", " target_size = (224,224),\n", " seed = 42,\n", " shuffle = True)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.771346Z", "iopub.status.idle": "2023-05-31T11:51:04.772100Z", "shell.execute_reply": "2023-05-31T11:51:04.771888Z", "shell.execute_reply.started": "2023-05-31T11:51:04.771865Z" } }, "outputs": [], "source": [ "validate_set = train_datagen.flow_from_dataframe(dataframe = train_labels,\n", " directory = train_path,\n", " x_col = \"id\",\n", " y_col = \"breed\",\n", " batch_size = 16,\n", " subset=\"validation\",\n", " class_mode = \"categorical\",\n", " target_size = (224,224),\n", " seed = 42,\n", " shuffle = True)\n" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.773456Z", "iopub.status.idle": "2023-05-31T11:51:04.774202Z", "shell.execute_reply": "2023-05-31T11:51:04.773990Z", "shell.execute_reply.started": "2023-05-31T11:51:04.773966Z" } }, "outputs": [], "source": [ "train_set.batch_size" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.775557Z", "iopub.status.idle": "2023-05-31T11:51:04.776315Z", "shell.execute_reply": "2023-05-31T11:51:04.776080Z", "shell.execute_reply.started": "2023-05-31T11:51:04.776058Z" } }, "outputs": [], "source": [ "resnet = ResNet50V2(input_shape = [224,224,3], weights = 'imagenet', include_top = False)\n", "\n", "for layer in resnet.layers:\n", " layer.trainable = False\n", "\n", "x = keras.layers.Flatten()(resnet.output)\n", "\n", "x = keras.layers.Dropout(0.4)(x)\n", "\n", "pred = keras.layers.Dense(10, activation='softmax')(x)\n", "\n", "model = tf.keras.models.Model(inputs=resnet.input, outputs=pred)\n", "\n", "opt = tf.keras.optimizers.Adam(learning_rate = 1e-5)\n", "model.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['accuracy'])\n", "\n", "train_step = train_set.n//train_set.batch_size\n", "validate_step = validate_set.n//validate_set.batch_size\n", "\n", "resnet50 = model.fit(train_set,validation_data = validate_set,epochs = 30,steps_per_epoch = train_step, validation_steps = validate_step)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.777684Z", "iopub.status.idle": "2023-05-31T11:51:04.778473Z", "shell.execute_reply": "2023-05-31T11:51:04.778252Z", "shell.execute_reply.started": "2023-05-31T11:51:04.778229Z" } }, "outputs": [], "source": [ "X_train_and_val, X_test, Y_train_and_val, Y_test = train_test_split(X_data, Y_data, test_size = 0.1)\n", "\n", "X_train, X_val, Y_train, Y_val = train_test_split(X_train_and_val, Y_train_and_val, test_size = 0.2)" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.779816Z", "iopub.status.idle": "2023-05-31T11:51:04.780606Z", "shell.execute_reply": "2023-05-31T11:51:04.780386Z", "shell.execute_reply.started": "2023-05-31T11:51:04.780355Z" } }, "outputs": [], "source": [ "epochs = 50\n", "batch_size = 62\n", "\n", "history = model.fit(X_train, Y_train, batch_size = batch_size, epochs = epochs, validation_data = (X_val, Y_val))" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.781954Z", "iopub.status.idle": "2023-05-31T11:51:04.782718Z", "shell.execute_reply": "2023-05-31T11:51:04.782506Z", "shell.execute_reply.started": "2023-05-31T11:51:04.782484Z" } }, "outputs": [], "source": [ "Y_pred = model.predict(X_test)\n", "score = model.evaluate(X_test, Y_test)\n", "print('Accuracy over the test set: \\n ', round((score[1]*100), 2), '%')" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.784137Z", "iopub.status.idle": "2023-05-31T11:51:04.784918Z", "shell.execute_reply": "2023-05-31T11:51:04.784705Z", "shell.execute_reply.started": "2023-05-31T11:51:04.784681Z" } }, "outputs": [], "source": [ "plt.figure(figsize=(12, 5))\n", "plt.plot(history.history['accuracy'])\n", "plt.plot(history.history['val_accuracy'])\n", "plt.title('Model Accuracy')\n", "plt.ylabel('Accuracy')\n", "plt.xlabel('Epochs')\n", "plt.legend(['train', 'val'])\n", "\n", "plt.show()" ] }, { "cell_type": "code", "execution_count": null, "metadata": { "execution": { "iopub.status.busy": "2023-05-31T11:51:04.786277Z", "iopub.status.idle": "2023-05-31T11:51:04.787025Z", "shell.execute_reply": "2023-05-31T11:51:04.786809Z", "shell.execute_reply.started": "2023-05-31T11:51:04.786787Z" } }, "outputs": [], "source": [ "plt.imshow(X_test[1,:,:,:])\n", "plt.show()\n", "\n", "# Finding max value from predition list and comaparing original value vs predicted\n", "print(\"Originally : \",labels['breed'][np.argmax(Y_test[1])])\n", "print(\"Predicted : \",labels['breed'][np.argmax(Y_pred[1])])" ] } ], "metadata": { "kernelspec": { "display_name": "Python 3", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.10.12" } }, "nbformat": 4, "nbformat_minor": 4 }