plant-diseases-detection.ipynb

{"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":"import os\nimport random\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\n\n#Visualization\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\nfrom sklearn import metrics\nfrom sklearn.metrics import accuracy_score, f1_score, recall_score, precision_score\nfrom sklearn.utils import class_weight\n\n#Tensorflow\nimport tensorflow as tf\nimport tensorflow.keras.layers as lyrs\nfrom tensorflow.keras.applications import ResNet50V2\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping\nimport tensorflow_hub as hub\nfrom tensorflow.keras.optimizers.legacy import Adam\n\nroot=\"/kaggle/input/new-plant-diseases-dataset/New Plant Diseases Dataset(Augmented)/New Plant Diseases Dataset(Augmented)\"\nIMG_SIZE=224","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-04-14T18:57:19.669341Z","iopub.execute_input":"2023-04-14T18:57:19.669812Z","iopub.status.idle":"2023-04-14T18:57:19.677721Z","shell.execute_reply.started":"2023-04-14T18:57:19.669739Z","shell.execute_reply":"2023-04-14T18:57:19.676621Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def supervised_metrics(y_true, y_pred):\n    \"\"\"Meterics for a Supervised Learning model:\"\"\"\n    print(\"Accuracy : {} %\".format(accuracy_score(y_true, y_pred)*100))\n    print(\"F1 Score : {}\".format(f1_score(y_true, y_pred, average='weighted')))\n    print(\"Recall : {}\".format(recall_score(y_true, y_pred, average='weighted')))\n    print(\"Precision : {}\".format(precision_score(y_true, y_pred, average='weighted')))\n    \ndef view_random_image(root_path,folder,class_folder):\n    path=root_path+'/'+folder+'/'+class_folder\n    rand=random.choice(os.listdir(path))\n    random_image=mpimg.imread(path+'/'+rand)\n    plt.imshow(random_image)\n    plt.title(\"File Name: \" + rand)\n    \ndef pre_process_image(path, image_shape=224, channels=3, norm_factor=255.):\n    '''Pre-Processing the Image before sending it to the model'''\n    img = tf.io.read_file(path)\n    img = tf.image.decode_image(img, channels=channels)\n    img = tf.image.resize(img, size = (image_shape, image_shape))\n    img = tf.expand_dims(img, axis=0)\n    img = img/norm_factor\n    return img\n\ndef random_tester(root_path, classes, model, class_type=\"binary\"):\n    '''Random Class Folder Selection'''\n    path=root_path\n    class_folder=random.choice(os.listdir(path))\n    \n    '''Random File Selection'''\n    folder_path=path+'/'+class_folder+'/'\n    rand=random.choice(os.listdir(folder_path))\n    file_path=folder_path+'/'+rand\n    random_image=mpimg.imread(file_path)\n    \n    '''Prediction'''\n    predicted_value=model.predict(pre_process_image(file_path)) \n    if(class_type==\"binary\"):\n        predicted_label=classes[custom_rounder(predicted_value)]\n    else:\n        index=tf.math.round(predicted_value).numpy()\n        index=np.argmax(index)\n        predicted_label=classes[index]\n        \n    '''Visualize'''\n    plt.imshow(random_image)\n    plt.title(\"Prediction:\" + predicted_label +\"\\n\" +\"True class: \"+ class_folder)\n    plt.show()\n    \ndef loss_curve_plot(df):\n    \"\"\" Dataframe (df) is history of the fit of the NN model\n    The df consists of train and validation fit data\n    \"\"\"\n    history = df.history\n    val_accuracy = history[\"val_accuracy\"]\n    val_loss = history[\"val_loss\"]\n    train_accuracy = history[\"accuracy\"]\n    train_loss = history[\"loss\"]\n    \n    \"\"\"Accuracy Plot\"\"\"\n    plt.plot(train_accuracy, label=\"Train Accuracy\")\n    plt.plot(val_accuracy, label=\"Validation Accuracy\")\n    plt.title(\"Accuracy Curves\")\n    plt.xlabel('epoch')\n    plt.ylabel('accuracy')\n    plt.legend()\n    plt.show()\n    \n    \"\"\"Loss Plot\"\"\"\n    plt.plot(train_loss, label=\"Train loss\")\n    plt.plot(val_loss, label=\"Validation loss\")\n    plt.title(\"Loss Curves\")\n    plt.xlabel('epoch')\n    plt.ylabel('loss')\n    plt.legend()\n    plt.show()\n    \ndef confusion_matrix_plot(y_true, y_pred, figsize=(30,30)):\n    \"\"\"\"Confusion Matrix for true values and predicted values\"\"\"\n    cm = metrics.confusion_matrix(y_true, y_pred)\n    cm = cm.astype('float') / cm.sum(axis=1)\n    plt.figure(figsize = figsize)\n    sns.heatmap(cm, annot=True, cmap=\"crest\")","metadata":{"execution":{"iopub.status.busy":"2023-04-14T18:57:24.762103Z","iopub.execute_input":"2023-04-14T18:57:24.762505Z","iopub.status.idle":"2023-04-14T18:57:24.781438Z","shell.execute_reply.started":"2023-04-14T18:57:24.762470Z","shell.execute_reply":"2023-04-14T18:57:24.780026Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"augmentation=ImageDataGenerator(\n    rescale=1/225.,\n    rotation_range=0.3,\n    vertical_flip=True,\n    horizontal_flip=True\n)\n\ntrain_datagen=augmentation.flow_from_directory(\n    root+\"/train\",\n    target_size=(IMG_SIZE, IMG_SIZE),\n    batch_size=32,\n    class_mode='categorical',\n    shuffle=True\n)\n\ntest_datagen=augmentation.flow_from_directory(\n    root+\"/valid\",\n    target_size=(IMG_SIZE, IMG_SIZE),\n    batch_size=32,\n    class_mode='categorical',\n    shuffle=False\n)","metadata":{"execution":{"iopub.status.busy":"2023-04-14T18:57:29.586396Z","iopub.execute_input":"2023-04-14T18:57:29.586786Z","iopub.status.idle":"2023-04-14T18:57:39.611186Z","shell.execute_reply.started":"2023-04-14T18:57:29.586736Z","shell.execute_reply":"2023-04-14T18:57:39.610110Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model_base =ResNet50V2(\n    input_shape=(224,224,3),\n    include_top=False,\n)\nmodel_base.trainable = False\n\n#Transfer Learning Model\ninputs=tf.keras.Input(shape=(224,224,3))\nx=model_base(inputs)\nx=lyrs.GlobalAveragePooling2D()(x)\nx= lyrs.BatchNormalization()(x)\nx= lyrs.Dropout(0.5)(x)\noutputs=lyrs.Dense(38, activation=\"softmax\")(x)\n\n\nmodel = tf.keras.Model(inputs=inputs, outputs=outputs)\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2023-04-14T18:57:53.855977Z","iopub.execute_input":"2023-04-14T18:57:53.856921Z","iopub.status.idle":"2023-04-14T18:57:56.007716Z","shell.execute_reply.started":"2023-04-14T18:57:53.856870Z","shell.execute_reply":"2023-04-14T18:57:56.006627Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"checkpointer = ModelCheckpoint('rice_leaf.hdf5',verbose=1, save_best_only= True)\nearly_stopping = EarlyStopping(monitor= 'val_loss', patience= 3)\n\nmodel.compile(loss=\"categorical_crossentropy\", optimizer=Adam(lr=1e-3, decay=1e-3), metrics=[\"accuracy\"])\nmodel_history=model.fit(x=train_datagen, \n         steps_per_epoch=32,\n         validation_data=test_datagen,\n         epochs=50,\n         callbacks=[checkpointer, early_stopping])","metadata":{"execution":{"iopub.status.busy":"2023-04-14T18:58:00.917124Z","iopub.execute_input":"2023-04-14T18:58:00.917716Z","iopub.status.idle":"2023-04-14T21:51:50.891108Z","shell.execute_reply.started":"2023-04-14T18:58:00.917680Z","shell.execute_reply":"2023-04-14T21:51:50.889953Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"labels=test_datagen.classes\nmodel=tf.keras.models.load_model(\"/kaggle/working/rice_leaf.hdf5\")\ny_pred=model.predict(test_datagen)\nprediction=tf.math.round(y_pred).numpy()\nprediction=prediction.argmax(axis=1)\n\nsupervised_metrics(labels, prediction)","metadata":{"execution":{"iopub.status.busy":"2023-04-14T21:53:03.206478Z","iopub.execute_input":"2023-04-14T21:53:03.206920Z","iopub.status.idle":"2023-04-14T21:57:02.652995Z","shell.execute_reply.started":"2023-04-14T21:53:03.206882Z","shell.execute_reply":"2023-04-14T21:57:02.651824Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"confusion_matrix_plot(labels, prediction, figsize=(10,10))","metadata":{"execution":{"iopub.status.busy":"2023-04-13T13:17:25.020582Z","iopub.execute_input":"2023-04-13T13:17:25.021677Z","iopub.status.idle":"2023-04-13T13:17:30.111860Z","shell.execute_reply.started":"2023-04-13T13:17:25.021621Z","shell.execute_reply":"2023-04-13T13:17:30.110927Z"},"trusted":true},"execution_count":null,"outputs":[]}]}