{"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":"# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python Docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load\n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the read-only \"../input/\" directory\n# For example, running this (by clicking run or pressing Shift+Enter) will list all files under the input directory\n\nimport os\nfor dirname, _, filenames in os.walk('/kaggle/input'):\n    for filename in filenames:\n        print(os.path.join(dirname, filename))\n\n# You can write up to 20GB to the current directory (/kaggle/working/) that gets preserved as output when you create a version using \"Save & Run All\" \n# You can also write temporary files to /kaggle/temp/, but they won't be saved outside of the current session","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_kg_hide-output":true,"_kg_hide-input":true,"execution":{"iopub.status.busy":"2022-09-11T11:23:43.773501Z","iopub.execute_input":"2022-09-11T11:23:43.773863Z","iopub.status.idle":"2022-09-11T11:24:01.178857Z","shell.execute_reply.started":"2022-09-11T11:23:43.773830Z","shell.execute_reply":"2022-09-11T11:24:01.177880Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nimport datetime\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.metrics import accuracy_score\nimport tensorflow as tf\nfrom tensorflow.keras import models, layers\nfrom tensorflow.keras.preprocessing import image\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator\nfrom tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\nfrom tensorflow.keras.applications import EfficientNetB0\nfrom tensorflow.keras.optimizers import Adam\n\n# ignoring warnings\nimport warnings\nwarnings.simplefilter(\"ignore\")\n\nimport os, cv2, json\nfrom PIL import Image","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:01.180372Z","iopub.execute_input":"2022-09-11T11:24:01.180671Z","iopub.status.idle":"2022-09-11T11:24:07.792852Z","shell.execute_reply.started":"2022-09-11T11:24:01.180642Z","shell.execute_reply":"2022-09-11T11:24:07.791881Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"WORK_DIR = '../input/cassava-leaf-disease-classification'\nos.listdir(WORK_DIR)","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:07.794069Z","iopub.execute_input":"2022-09-11T11:24:07.794713Z","iopub.status.idle":"2022-09-11T11:24:07.808456Z","shell.execute_reply.started":"2022-09-11T11:24:07.794673Z","shell.execute_reply":"2022-09-11T11:24:07.806594Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print('Train images: %d' %len(os.listdir(\n    os.path.join(WORK_DIR, \"train_images\"))))","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:07.810231Z","iopub.execute_input":"2022-09-11T11:24:07.810802Z","iopub.status.idle":"2022-09-11T11:24:07.831699Z","shell.execute_reply.started":"2022-09-11T11:24:07.810766Z","shell.execute_reply":"2022-09-11T11:24:07.830678Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with open(os.path.join(WORK_DIR, \"label_num_to_disease_map.json\")) as file:\n    print(json.dumps(json.loads(file.read()), indent=4))","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:07.834709Z","iopub.execute_input":"2022-09-11T11:24:07.835131Z","iopub.status.idle":"2022-09-11T11:24:07.844072Z","shell.execute_reply.started":"2022-09-11T11:24:07.835103Z","shell.execute_reply":"2022-09-11T11:24:07.842722Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_labels = pd.read_csv(os.path.join(WORK_DIR, \"train.csv\"))\ntrain_labels.head()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:07.845486Z","iopub.execute_input":"2022-09-11T11:24:07.845981Z","iopub.status.idle":"2022-09-11T11:24:07.885927Z","shell.execute_reply.started":"2022-09-11T11:24:07.845947Z","shell.execute_reply":"2022-09-11T11:24:07.885014Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sns.set_style(\"whitegrid\")\nfig, ax = plt.subplots(figsize = (6, 4))\n\nfor i in ['top', 'right', 'left']:\n    ax.spines[i].set_visible(False)\nax.spines['bottom'].set_color('black')\n\nsns.countplot(train_labels.label, edgecolor = 'black',\n              palette = reversed(sns.color_palette(\"viridis\", 5)))\nplt.xlabel('Classes', fontfamily = 'serif', size = 15)\nplt.ylabel('Count', fontfamily = 'serif', size = 15)\nplt.xticks(fontfamily = 'serif', size = 12)\nplt.yticks(fontfamily = 'serif', size = 12)\nax.grid(axis = 'y', linestyle = '--', alpha = 0.9)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:07.887158Z","iopub.execute_input":"2022-09-11T11:24:07.887585Z","iopub.status.idle":"2022-09-11T11:24:08.120780Z","shell.execute_reply.started":"2022-09-11T11:24:07.887551Z","shell.execute_reply":"2022-09-11T11:24:08.119920Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = train_labels[train_labels.label == 0].sample(3)\nplt.figure(figsize=(15, 5))\nfor ind, (image_id, label) in enumerate(zip(sample.image_id, sample.label)):\n    plt.subplot(1, 3, ind + 1)\n    img = cv2.imread(os.path.join(WORK_DIR, \"train_images\", image_id))\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    plt.imshow(img)\n    plt.axis(\"off\")\n    \nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:08.122222Z","iopub.execute_input":"2022-09-11T11:24:08.122548Z","iopub.status.idle":"2022-09-11T11:24:08.731800Z","shell.execute_reply.started":"2022-09-11T11:24:08.122515Z","shell.execute_reply":"2022-09-11T11:24:08.730966Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = train_labels[train_labels.label == 1].sample(3)\nplt.figure(figsize=(15, 5))\nfor ind, (image_id, label) in enumerate(zip(sample.image_id, sample.label)):\n    plt.subplot(1, 3, ind + 1)\n    img = cv2.imread(os.path.join(WORK_DIR, \"train_images\", image_id))\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    plt.imshow(img)\n    plt.axis(\"off\")\n    \nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:08.733278Z","iopub.execute_input":"2022-09-11T11:24:08.733779Z","iopub.status.idle":"2022-09-11T11:24:09.302632Z","shell.execute_reply.started":"2022-09-11T11:24:08.733746Z","shell.execute_reply":"2022-09-11T11:24:09.301780Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = train_labels[train_labels.label == 2].sample(3)\nplt.figure(figsize=(15, 5))\nfor ind, (image_id, label) in enumerate(zip(sample.image_id, sample.label)):\n    plt.subplot(1, 3, ind + 1)\n    img = cv2.imread(os.path.join(WORK_DIR, \"train_images\", image_id))\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    plt.imshow(img)\n    plt.axis(\"off\")\n    \nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:09.304007Z","iopub.execute_input":"2022-09-11T11:24:09.304530Z","iopub.status.idle":"2022-09-11T11:24:09.860702Z","shell.execute_reply.started":"2022-09-11T11:24:09.304496Z","shell.execute_reply":"2022-09-11T11:24:09.859792Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = train_labels[train_labels.label == 3].sample(3)\nplt.figure(figsize=(15, 5))\nfor ind, (image_id, label) in enumerate(zip(sample.image_id, sample.label)):\n    plt.subplot(1, 3, ind + 1)\n    img = cv2.imread(os.path.join(WORK_DIR, \"train_images\", image_id))\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    plt.imshow(img)\n    plt.axis(\"off\")\n    \nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:09.862118Z","iopub.execute_input":"2022-09-11T11:24:09.862721Z","iopub.status.idle":"2022-09-11T11:24:10.429555Z","shell.execute_reply.started":"2022-09-11T11:24:09.862682Z","shell.execute_reply":"2022-09-11T11:24:10.428720Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample = train_labels[train_labels.label == 4].sample(3)\nplt.figure(figsize=(15, 5))\nfor ind, (image_id, label) in enumerate(zip(sample.image_id, sample.label)):\n    plt.subplot(1, 3, ind + 1)\n    img = cv2.imread(os.path.join(WORK_DIR, \"train_images\", image_id))\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    plt.imshow(img)\n    plt.axis(\"off\")\n    \nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:10.430886Z","iopub.execute_input":"2022-09-11T11:24:10.431408Z","iopub.status.idle":"2022-09-11T11:24:10.990351Z","shell.execute_reply.started":"2022-09-11T11:24:10.431375Z","shell.execute_reply":"2022-09-11T11:24:10.989504Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Main parameters\nBATCH_SIZE = 8\nSTEPS_PER_EPOCH = len(train_labels)*0.8 / BATCH_SIZE\nVALIDATION_STEPS = len(train_labels)*0.2 / BATCH_SIZE\nEPOCHS = 20\nTARGET_SIZE = 512","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:10.991681Z","iopub.execute_input":"2022-09-11T11:24:10.992212Z","iopub.status.idle":"2022-09-11T11:24:10.997204Z","shell.execute_reply.started":"2022-09-11T11:24:10.992177Z","shell.execute_reply":"2022-09-11T11:24:10.996366Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_labels.label = train_labels.label.astype('str')\n\ntrain_datagen = ImageDataGenerator(validation_split = 0.2,\n                                     preprocessing_function = None,\n                                     rotation_range = 45,\n                                     zoom_range = 0.2,\n                                     horizontal_flip = True,\n                                     vertical_flip = True,\n                                     fill_mode = 'nearest',\n                                     shear_range = 0.1,\n                                     height_shift_range = 0.1,\n                                     width_shift_range = 0.1)\n\ntrain_generator = train_datagen.flow_from_dataframe(train_labels,\n                         directory = os.path.join(WORK_DIR, \"train_images\"),\n                         subset = \"training\",\n                         x_col = \"image_id\",\n                         y_col = \"label\",\n                         target_size = (TARGET_SIZE, TARGET_SIZE),\n                         batch_size = BATCH_SIZE,\n                         class_mode = \"sparse\")\n\n\nvalidation_datagen = ImageDataGenerator(validation_split = 0.2)\n\nvalidation_generator = validation_datagen.flow_from_dataframe(train_labels,\n                         directory = os.path.join(WORK_DIR, \"train_images\"),\n                         subset = \"validation\",\n                         x_col = \"image_id\",\n                         y_col = \"label\",\n                         target_size = (TARGET_SIZE, TARGET_SIZE),\n                         batch_size = BATCH_SIZE,\n                         class_mode = \"sparse\")","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:11.001616Z","iopub.execute_input":"2022-09-11T11:24:11.002282Z","iopub.status.idle":"2022-09-11T11:24:31.212487Z","shell.execute_reply.started":"2022-09-11T11:24:11.002232Z","shell.execute_reply":"2022-09-11T11:24:31.211435Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"img_path = os.path.join(WORK_DIR, \"train_images\", train_labels.image_id[20])\nimg = image.load_img(img_path, target_size = (TARGET_SIZE, TARGET_SIZE))\nimg_tensor = image.img_to_array(img)\nimg_tensor = np.expand_dims(img_tensor, axis = 0)\nimg_tensor /= 255.\n\nplt.imshow(img_tensor[0])\nplt.axis('off')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:31.214192Z","iopub.execute_input":"2022-09-11T11:24:31.214879Z","iopub.status.idle":"2022-09-11T11:24:31.462883Z","shell.execute_reply.started":"2022-09-11T11:24:31.214839Z","shell.execute_reply":"2022-09-11T11:24:31.461970Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"generator = train_datagen.flow_from_dataframe(train_labels.iloc[20:21],\n                         directory = os.path.join(WORK_DIR, \"train_images\"),\n                         x_col = \"image_id\",\n                         y_col = \"label\",\n                         target_size = (TARGET_SIZE, TARGET_SIZE),\n                         batch_size = BATCH_SIZE,\n                         class_mode = \"sparse\")\n\naug_images = [generator[0][0][0]/255 for i in range(10)]\nfig, axes = plt.subplots(2, 5, figsize = (20, 10))\naxes = axes.flatten()\nfor img, ax in zip(aug_images, axes):\n    ax.imshow(img)\n    ax.axis('off')\nplt.tight_layout()\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:31.464842Z","iopub.execute_input":"2022-09-11T11:24:31.465660Z","iopub.status.idle":"2022-09-11T11:24:33.947917Z","shell.execute_reply.started":"2022-09-11T11:24:31.465622Z","shell.execute_reply":"2022-09-11T11:24:33.947084Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def create_model():\n    conv_base = EfficientNetB0(include_top = False, weights = None,\n                               input_shape = (TARGET_SIZE, TARGET_SIZE, 3))\n    model = conv_base.output\n    model = layers.GlobalAveragePooling2D()(model)\n    model = layers.Dense(5, activation = \"softmax\")(model)\n    model = models.Model(conv_base.input, model)\n\n    model.compile(optimizer = Adam(lr = 0.001),\n                  loss = \"sparse_categorical_crossentropy\",\n                  metrics = [\"acc\"])\n    return model","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:33.949337Z","iopub.execute_input":"2022-09-11T11:24:33.949884Z","iopub.status.idle":"2022-09-11T11:24:33.956674Z","shell.execute_reply.started":"2022-09-11T11:24:33.949849Z","shell.execute_reply":"2022-09-11T11:24:33.955649Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = create_model()\nmodel.summary()","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:33.958112Z","iopub.execute_input":"2022-09-11T11:24:33.958668Z","iopub.status.idle":"2022-09-11T11:24:38.512731Z","shell.execute_reply.started":"2022-09-11T11:24:33.958634Z","shell.execute_reply":"2022-09-11T11:24:38.511787Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print('Our EfficientNet CNN has %d layers' %len(model.layers))","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:38.514218Z","iopub.execute_input":"2022-09-11T11:24:38.514821Z","iopub.status.idle":"2022-09-11T11:24:38.523097Z","shell.execute_reply.started":"2022-09-11T11:24:38.514781Z","shell.execute_reply":"2022-09-11T11:24:38.522103Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.load_weights('../input/cassava-leaf-disease-models/basic_EfNetB0_imagenet_512.h5')","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:38.524348Z","iopub.execute_input":"2022-09-11T11:24:38.524761Z","iopub.status.idle":"2022-09-11T11:24:39.405447Z","shell.execute_reply.started":"2022-09-11T11:24:38.524726Z","shell.execute_reply":"2022-09-11T11:24:39.404495Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model_save = ModelCheckpoint('./EffNetB0_512_8_best_weights.h5', \n                             save_best_only = True, \n                             save_weights_only = True,\n                             monitor = 'val_loss', \n                             mode = 'min', verbose = 1)\nearly_stop = EarlyStopping(monitor = 'val_loss', min_delta = 0.001, \n                           patience = 5, mode = 'min', verbose = 1,\n                           restore_best_weights = True)\nreduce_lr = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.3, \n                              patience = 2, min_delta = 0.001, \n                              mode = 'min', verbose = 1)\n\n\nhistory = model.fit(\n    train_generator,\n    steps_per_epoch = STEPS_PER_EPOCH,\n    epochs = EPOCHS,\n    validation_data = validation_generator,\n    validation_steps = VALIDATION_STEPS,\n    callbacks = [model_save, early_stop, reduce_lr]\n)","metadata":{"execution":{"iopub.status.busy":"2022-09-11T11:24:39.406814Z","iopub.execute_input":"2022-09-11T11:24:39.407193Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"acc = history.history['acc']\nval_acc = history.history['val_acc']\nloss = history.history['loss']\nval_loss = history.history['val_loss']\n\nepochs = range(1, len(acc) + 1)\n\nfig, (ax1, ax2) = plt.subplots(1, 2, figsize=(15, 5))\nsns.set_style(\"white\")\nplt.suptitle('Train history', size = 15)\n\nax1.plot(epochs, acc, \"bo\", label = \"Training acc\")\nax1.plot(epochs, val_acc, \"b\", label = \"Validation acc\")\nax1.set_title(\"Training and validation acc\")\nax1.legend()\n\nax2.plot(epochs, loss, \"bo\", label = \"Training loss\", color = 'red')\nax2.plot(epochs, val_loss, \"b\", label = \"Validation loss\", color = 'red')\nax2.set_title(\"Training and validation loss\")\nax2.legend()\n\nplt.show()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model.save('./EffNetB0_512_8.h5')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def activation_layer_vis(img, activation_layer = 0, layers = 10):\n    layer_outputs = [layer.output for layer in model.layers[:layers]]\n    activation_model = models.Model(inputs = model.input, outputs = layer_outputs)\n    activations = activation_model.predict(img)\n    \n    rows = int(activations[activation_layer].shape[3] / 3)\n    cols = int(activations[activation_layer].shape[3] / rows)\n    fig, axes = plt.subplots(rows, cols, figsize = (15, 15 * cols))\n    axes = axes.flatten()\n    \n    for i, ax in zip(range(activations[activation_layer].shape[3]), axes):\n        ax.matshow(activations[activation_layer][0, :, :, i], cmap = 'viridis')\n        ax.axis('off')\n    plt.tight_layout()\n    plt.show()","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"activation_layer_vis(img_tensor, 0)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def all_activations_vis(img, layers = 10):\n    layer_outputs = [layer.output for layer in model.layers[:layers]]\n    activation_model = models.Model(inputs = model.input, outputs = layer_outputs)\n    activations = activation_model.predict(img)\n    \n    layer_names = []\n    for layer in model.layers[:layers]: \n        layer_names.append(layer.name) \n\n    images_per_row = 3\n    for layer_name, layer_activation in zip(layer_names, activations): \n        n_features = layer_activation.shape[-1] \n\n        size = layer_activation.shape[1] \n\n        n_cols = n_features // images_per_row \n        display_grid = np.zeros((size * n_cols, images_per_row * size)) \n\n        for col in range(n_cols): \n            for row in range(images_per_row): \n                channel_image = layer_activation[0, :, :, col * images_per_row + row] \n                channel_image -= channel_image.mean() \n                channel_image /= channel_image.std() \n                channel_image *= 64 \n                channel_image += 128 \n                channel_image = np.clip(channel_image, 0, 255).astype('uint8') \n                display_grid[col * size : (col + 1) * size, \n                             row * size : (row + 1) * size] = channel_image \n        scale = 1. / size \n        plt.figure(figsize=(scale * 5 * display_grid.shape[1], \n                            scale * 5 * display_grid.shape[0])) \n        plt.title(layer_name) \n        plt.grid(False)\n        plt.axis('off')\n        plt.imshow(display_grid, aspect = 'auto', cmap = 'viridis')","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"all_activations_vis(img_tensor, 5)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}