{"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":"markdown","source":"<h1><center> 🍎 Classify foliar diseases in apple trees</center></h1>","metadata":{}},{"cell_type":"markdown","source":"# 1. Problem Statement ？\n\nApples are one of the most important temperate fruit crops in the world. Foliar (leaf) diseases pose a major threat to the overall productivity and quality of apple orchards. The current process for disease diagnosis in apple orchards is based on manual scouting by humans, which is time-consuming and expensive.\n\nThe main objective of the competition is to develop machine learning-based models to accurately classify a given leaf image from the test dataset to a particular disease category, and to identify an individual disease from multiple disease symptoms on a single leaf image.\n","metadata":{}},{"cell_type":"markdown","source":"## libraries ","metadata":{}},{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\n%matplotlib inline\nimport seaborn as sns\nimport matplotlib.pyplot as plt\nimport cv2\nimport os\nimport warnings\nwarnings.filterwarnings('ignore')\nimport tensorflow as tf\nimport random\nimport albumentations as A\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.layers import Dense,Activation,Flatten, Conv2D, MaxPooling2D\nfrom tensorflow.keras.models import Sequential\nfrom tensorflow.keras.callbacks import ModelCheckpoint,EarlyStopping","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:26.129946Z","iopub.execute_input":"2021-08-29T02:01:26.131959Z","iopub.status.idle":"2021-08-29T02:01:34.237494Z","shell.execute_reply.started":"2021-08-29T02:01:26.131797Z","shell.execute_reply":"2021-08-29T02:01:34.236407Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# 2. About Dataset","metadata":{}},{"cell_type":"code","source":"train_image_path = '../input/plant-pathology-2021-fgvc8/train_images'\ntest_image_path = '../input/plant-pathology-2021-fgvc8/test_images'\ntrain_df_path = '../input/plant-pathology-2021-fgvc8/train.csv'\ntest_df_path = '../input/plant-pathology-2021-fgvc8/sample_submission.csv'","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:34.238814Z","iopub.execute_input":"2021-08-29T02:01:34.239106Z","iopub.status.idle":"2021-08-29T02:01:34.243665Z","shell.execute_reply.started":"2021-08-29T02:01:34.239079Z","shell.execute_reply":"2021-08-29T02:01:34.242590Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"> 📌**Note**:\n* `train.csv` contains information about the image files available in `train_images`. It contains 18632 rows(images) with 2 columns i.e (image , labels )\n* `test.csv` The test set images. This competition has a hidden test set: only three images are provided here as samples while the remaining 5,000 images will be available to your notebook once it is submitted.","metadata":{}},{"cell_type":"code","source":"df_train = pd.read_csv(train_df_path)\ndf_test=pd.read_csv(test_df_path)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:34.245385Z","iopub.execute_input":"2021-08-29T02:01:34.245677Z","iopub.status.idle":"2021-08-29T02:01:34.294353Z","shell.execute_reply.started":"2021-08-29T02:01:34.245648Z","shell.execute_reply":"2021-08-29T02:01:34.293386Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_test","metadata":{"execution":{"iopub.status.busy":"2021-08-29T04:33:42.984399Z","iopub.execute_input":"2021-08-29T04:33:42.984827Z","iopub.status.idle":"2021-08-29T04:33:43.028309Z","shell.execute_reply.started":"2021-08-29T04:33:42.984783Z","shell.execute_reply":"2021-08-29T04:33:43.027312Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_train.labels.value_counts()","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:34.298302Z","iopub.execute_input":"2021-08-29T02:01:34.298605Z","iopub.status.idle":"2021-08-29T02:01:34.319776Z","shell.execute_reply.started":"2021-08-29T02:01:34.298577Z","shell.execute_reply":"2021-08-29T02:01:34.318775Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"plt.figure(figsize=(15,12))\nlabels = sns.barplot(df_train.labels.value_counts().index,df_train.labels.value_counts())\nfor item in labels.get_xticklabels():\n    item.set_rotation(45)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:34.321129Z","iopub.execute_input":"2021-08-29T02:01:34.321437Z","iopub.status.idle":"2021-08-29T02:01:34.648079Z","shell.execute_reply.started":"2021-08-29T02:01:34.321390Z","shell.execute_reply":"2021-08-29T02:01:34.646994Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"> 📌**Note**:\n* We have multiple labels for eg. label can be **scab** or **scab and rust**\n* Main labels are - **scab** , **healthy** , **frog_eye_leaf_spot** , **rust** , **complex** and **powdery_mildew**","metadata":{}},{"cell_type":"markdown","source":"## Batch Visualisation of Images ","metadata":{}},{"cell_type":"code","source":"def batch_visualize(df,batch_size,path):\n    sample_df = df_train.sample(9)\n    image_names = sample_df[\"image\"].values\n    labels = sample_df[\"labels\"].values\n    plt.figure(figsize=(16, 12))\n    \n    for image_ind, (image_name, label) in enumerate(zip(image_names, labels)):\n        plt.subplot(3, 3, image_ind + 1)\n        image = cv2.imread(os.path.join(path, image_name))\n        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n        plt.imshow(image)\n        plt.title(f\"{label}\", fontsize=12)\n        plt.axis(\"off\")\n    plt.show()\n    \nbatch_visualize(df_train,9,train_image_path)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:34.649511Z","iopub.execute_input":"2021-08-29T02:01:34.649890Z","iopub.status.idle":"2021-08-29T02:01:45.027938Z","shell.execute_reply.started":"2021-08-29T02:01:34.649847Z","shell.execute_reply":"2021-08-29T02:01:45.027075Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Batch visualisation with labels","metadata":{}},{"cell_type":"code","source":"def batch_visualize_with_label(df,batch_size,path,label): \n    sample_df = df_train[df_train[\"labels\"]==label].sample(9)\n    image_names = sample_df[\"image\"].values\n    labels = sample_df[\"labels\"].values\n    plt.figure(figsize=(16, 12))\n    \n    for image_ind, (image_name, label) in enumerate(zip(image_names, labels)):\n        plt.subplot(3, 3, image_ind + 1)\n        image = cv2.imread(os.path.join(path, image_name))\n        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n        plt.imshow(image)\n        plt.axis(\"off\")\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:45.029018Z","iopub.execute_input":"2021-08-29T02:01:45.029425Z","iopub.status.idle":"2021-08-29T02:01:45.036033Z","shell.execute_reply.started":"2021-08-29T02:01:45.029395Z","shell.execute_reply":"2021-08-29T02:01:45.035150Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise healthy leaves","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'healthy')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:45.038237Z","iopub.execute_input":"2021-08-29T02:01:45.038683Z","iopub.status.idle":"2021-08-29T02:01:55.854938Z","shell.execute_reply.started":"2021-08-29T02:01:45.038636Z","shell.execute_reply":"2021-08-29T02:01:55.853975Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise scab leaves ","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'scab')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:01:55.856810Z","iopub.execute_input":"2021-08-29T02:01:55.857119Z","iopub.status.idle":"2021-08-29T02:02:06.887892Z","shell.execute_reply.started":"2021-08-29T02:01:55.857088Z","shell.execute_reply":"2021-08-29T02:02:06.886758Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise frog_eye_leaf_spot  leaves","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'frog_eye_leaf_spot')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:06.889429Z","iopub.execute_input":"2021-08-29T02:02:06.889808Z","iopub.status.idle":"2021-08-29T02:02:17.280555Z","shell.execute_reply.started":"2021-08-29T02:02:06.889771Z","shell.execute_reply":"2021-08-29T02:02:17.279619Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise rust leaves ","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'rust')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:17.281766Z","iopub.execute_input":"2021-08-29T02:02:17.282092Z","iopub.status.idle":"2021-08-29T02:02:23.988234Z","shell.execute_reply.started":"2021-08-29T02:02:17.282062Z","shell.execute_reply":"2021-08-29T02:02:23.987296Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise complex leaves","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'complex')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:23.989371Z","iopub.execute_input":"2021-08-29T02:02:23.989803Z","iopub.status.idle":"2021-08-29T02:02:34.139261Z","shell.execute_reply.started":"2021-08-29T02:02:23.989771Z","shell.execute_reply":"2021-08-29T02:02:34.138272Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### Visualise powdery_mildew leaves","metadata":{}},{"cell_type":"code","source":"batch_visualize_with_label(df_train,9,train_image_path,'powdery_mildew')","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:34.140597Z","iopub.execute_input":"2021-08-29T02:02:34.140946Z","iopub.status.idle":"2021-08-29T02:02:44.882320Z","shell.execute_reply.started":"2021-08-29T02:02:34.140915Z","shell.execute_reply":"2021-08-29T02:02:44.881388Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# 3. Tensorflow Dataset Generation","metadata":{}},{"cell_type":"code","source":"HEIGHT = 128\nWIDTH=128\nSEED = 45\nBATCH_SIZE= 64\n\ntrain_datagen = ImageDataGenerator(rescale = 1/255.,\n    rotation_range=20,\n    width_shift_range=0.2,\n    height_shift_range=0.2,\n    horizontal_flip=True,\n    validation_split = 0.2,\n    zoom_range = 0.2,\n    shear_range = 0.2,\n    vertical_flip = False)\n\ntrain_dataset = train_datagen.flow_from_dataframe(\n    df_train,\n    directory = train_image_path,\n    x_col = \"image\",\n    y_col = \"labels\",\n    target_size = (HEIGHT,WIDTH),\n    class_mode='categorical',\n    batch_size = BATCH_SIZE,\n    subset = \"training\",\n    shuffle = True,\n    seed = SEED,\n    validate_filenames = False\n)\n\n\nvalidation_dataset = train_datagen.flow_from_dataframe(\n    df_train,\n    directory = train_image_path,\n    x_col = \"image\",\n    y_col = \"labels\",\n    target_size = (HEIGHT,WIDTH),\n    class_mode='categorical',\n    batch_size = BATCH_SIZE,\n    subset = \"validation\",\n    shuffle = True,\n    seed = SEED,\n    validate_filenames = False\n)\n\ntest_datagen = ImageDataGenerator(\n    rescale = 1./255\n)\nINPUT_SIZE = (HEIGHT,WIDTH,3)\ntest_dataset=test_datagen.flow_from_dataframe(\n    df_test,\n    directory=test_image_path,\n    x_col='image',\n    y_col=None,\n    class_mode=None,\n    target_size=INPUT_SIZE[:2]\n)\n","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:44.883544Z","iopub.execute_input":"2021-08-29T02:02:44.883821Z","iopub.status.idle":"2021-08-29T02:02:45.033977Z","shell.execute_reply.started":"2021-08-29T02:02:44.883794Z","shell.execute_reply":"2021-08-29T02:02:45.032881Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"> Refer Tensorflow docs for more information [here](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator)","metadata":{}},{"cell_type":"markdown","source":"# 4. Convolutional Neural Networks ","metadata":{}},{"cell_type":"code","source":"model=Sequential()\nmodel.add(Conv2D(32,(3,3),activation='relu',padding='same',input_shape=(HEIGHT,WIDTH,3)))\nmodel.add(MaxPooling2D(2,2))\nmodel.add(Conv2D(64,(3,3),activation='relu',padding='same'))\nmodel.add(MaxPooling2D(2,2))\nmodel.add(Conv2D(64,(3,3),activation='relu',padding='same'))\nmodel.add(MaxPooling2D(2,2))\nmodel.add(Conv2D(128,(3,3),activation='relu',padding='same'))\nmodel.add(MaxPooling2D(2,2))\nmodel.add(Flatten())\nmodel.add(Dense(12,activation='softmax'))","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:45.035334Z","iopub.execute_input":"2021-08-29T02:02:45.035623Z","iopub.status.idle":"2021-08-29T02:02:45.189091Z","shell.execute_reply.started":"2021-08-29T02:02:45.035595Z","shell.execute_reply":"2021-08-29T02:02:45.188055Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001),\n    loss='categorical_crossentropy',\n    metrics=['accuracy'])\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:45.190363Z","iopub.execute_input":"2021-08-29T02:02:45.190719Z","iopub.status.idle":"2021-08-29T02:02:45.212282Z","shell.execute_reply.started":"2021-08-29T02:02:45.190688Z","shell.execute_reply":"2021-08-29T02:02:45.211113Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"checkpoint_path = \"training_1/cp.ckpt\"\ncheckpoint_dir = os.path.dirname(checkpoint_path)\n\n# Create a callback that saves the model's weights\ncp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,\n                                                 save_weights_only=True,\n                                                 verbose=1)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:45.213491Z","iopub.execute_input":"2021-08-29T02:02:45.213758Z","iopub.status.idle":"2021-08-29T02:02:45.221933Z","shell.execute_reply.started":"2021-08-29T02:02:45.213732Z","shell.execute_reply":"2021-08-29T02:02:45.220951Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model_history=model.fit_generator(train_dataset,\n                                  validation_data=validation_dataset,\n                                  epochs=5,\n                                  steps_per_epoch=train_dataset.samples//128,\n                                 validation_steps=validation_dataset.samples//128,\n                                 callbacks=[cp_callback]\n                                 )","metadata":{"execution":{"iopub.status.busy":"2021-08-29T02:02:45.223101Z","iopub.execute_input":"2021-08-29T02:02:45.223399Z","iopub.status.idle":"2021-08-29T04:22:58.203984Z","shell.execute_reply.started":"2021-08-29T02:02:45.223345Z","shell.execute_reply":"2021-08-29T04:22:58.198975Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Prediction","metadata":{}},{"cell_type":"code","source":"train_dataset.class_indices.items()","metadata":{"execution":{"iopub.status.busy":"2021-08-29T04:22:58.210673Z","iopub.execute_input":"2021-08-29T04:22:58.211082Z","iopub.status.idle":"2021-08-29T04:22:58.225036Z","shell.execute_reply.started":"2021-08-29T04:22:58.211025Z","shell.execute_reply":"2021-08-29T04:22:58.224217Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"preds = model.predict(test_dataset)\nprint(preds)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T04:22:58.226203Z","iopub.execute_input":"2021-08-29T04:22:58.226694Z","iopub.status.idle":"2021-08-29T04:22:59.750428Z","shell.execute_reply.started":"2021-08-29T04:22:58.226656Z","shell.execute_reply":"2021-08-29T04:22:59.749462Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"preds_disease_ind=np.argmax(preds, axis=-1)","metadata":{"execution":{"iopub.status.busy":"2021-08-29T04:22:59.751831Z","iopub.execute_input":"2021-08-29T04:22:59.752157Z","iopub.status.idle":"2021-08-29T04:22:59.758131Z","shell.execute_reply.started":"2021-08-29T04:22:59.752128Z","shell.execute_reply":"2021-08-29T04:22:59.757099Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"preds_disease_ind","metadata":{"execution":{"iopub.status.busy":"2021-08-29T04:22:59.759533Z","iopub.execute_input":"2021-08-29T04:22:59.759865Z","iopub.status.idle":"2021-08-29T04:22:59.771499Z","shell.execute_reply.started":"2021-08-29T04:22:59.759819Z","shell.execute_reply":"2021-08-29T04:22:59.770778Z"},"trusted":true},"execution_count":null,"outputs":[]}]}