{"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":"3.6.6"},"kaggle":{"accelerator":"gpu","dataSources":[{"sourceId":13032,"databundleVersionId":862545,"sourceType":"competition"},{"sourceId":8619411,"sourceType":"datasetVersion","datasetId":5159316}],"dockerImageVersionId":25160,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":true}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"Welcome to the world where fashion meets computer vision! This is a starter kernel that applies Mask R-CNN with COCO pretrained weights to the task of [iMaterialist (Fashion) 2019 at FGVC6](https://www.kaggle.com/c/imaterialist-fashion-2019-FGVC6).","metadata":{}},{"cell_type":"code","source":"import os\nimport gc\nimport sys\nimport json\nimport glob\nimport random\nfrom pathlib import Path\n\nimport cv2\nimport numpy as np\nimport pandas as pd\nimport matplotlib.pyplot as plt\n\nimport itertools\nfrom tqdm import tqdm\n\nfrom imgaug import augmenters as iaa\nfrom sklearn.model_selection import StratifiedKFold, KFold","metadata":{"execution":{"iopub.status.busy":"2024-06-06T05:51:18.558385Z","iopub.execute_input":"2024-06-06T05:51:18.558684Z","iopub.status.idle":"2024-06-06T05:51:18.564336Z","shell.execute_reply.started":"2024-06-06T05:51:18.558636Z","shell.execute_reply":"2024-06-06T05:51:18.563424Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"DATA_DIR = Path('/kaggle/input/imaterialist-fashion-2019-FGVC6')\nROOT_DIR = Path('/kaggle/working')\n\n# For demonstration purpose, the classification ignores attributes (only categories),\n# and the image size is set to 512, which is the same as the size of submission masks\nNUM_CATS = 46\nIMAGE_SIZE = 512","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:00:54.240807Z","iopub.execute_input":"2024-06-06T06:00:54.241112Z","iopub.status.idle":"2024-06-06T06:00:54.245459Z","shell.execute_reply.started":"2024-06-06T06:00:54.241068Z","shell.execute_reply":"2024-06-06T06:00:54.244557Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Dowload Libraries and Pretrained Weights","metadata":{}},{"cell_type":"code","source":"!git clone https://www.github.com/matterport/Mask_RCNN.git\nos.chdir('Mask_RCNN')\n\n!rm -rf .git # to prevent an error when the kernel is committed\n!rm -rf images assets # to prevent displaying images at the bottom of a kernel","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","execution":{"iopub.status.busy":"2024-06-06T06:00:57.382785Z","iopub.execute_input":"2024-06-06T06:00:57.383110Z","iopub.status.idle":"2024-06-06T06:01:08.539538Z","shell.execute_reply.started":"2024-06-06T06:00:57.383056Z","shell.execute_reply":"2024-06-06T06:01:08.538573Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sys.path.append(ROOT_DIR/'Mask_RCNN')\nfrom mrcnn.config import Config\nfrom mrcnn import utils\nimport mrcnn.model as modellib\nfrom mrcnn import visualize\nfrom mrcnn.model import log","metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","execution":{"iopub.status.busy":"2024-06-06T06:01:08.541130Z","iopub.execute_input":"2024-06-06T06:01:08.541399Z","iopub.status.idle":"2024-06-06T06:01:08.547101Z","shell.execute_reply.started":"2024-06-06T06:01:08.541352Z","shell.execute_reply":"2024-06-06T06:01:08.546129Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!wget --quiet https://github.com/matterport/Mask_RCNN/releases/download/v2.0/mask_rcnn_coco.h5\n!ls -lh mask_rcnn_coco.h5\n\nCOCO_WEIGHTS_PATH = 'mask_rcnn_coco.h5'","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:01:11.072534Z","iopub.execute_input":"2024-06-06T06:01:11.072811Z","iopub.status.idle":"2024-06-06T06:03:27.311413Z","shell.execute_reply.started":"2024-06-06T06:01:11.072770Z","shell.execute_reply":"2024-06-06T06:03:27.310196Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Set Config","metadata":{}},{"cell_type":"markdown","source":"Mask R-CNN has a load of hyperparameters. I only adjust some of them.","metadata":{}},{"cell_type":"code","source":"class FashionConfig(Config):\n    NAME = \"fashion\"\n    NUM_CLASSES = NUM_CATS + 1 # +1 for the background class\n    \n    GPU_COUNT = 1\n    IMAGES_PER_GPU = 4 # a memory error occurs when IMAGES_PER_GPU is too high\n    \n    BACKBONE = 'resnet50'\n    \n    IMAGE_MIN_DIM = IMAGE_SIZE\n    IMAGE_MAX_DIM = IMAGE_SIZE    \n    IMAGE_RESIZE_MODE = 'none'\n    \n    RPN_ANCHOR_SCALES = (16, 32, 64, 128, 256)\n    #DETECTION_NMS_THRESHOLD = 0.0\n    \n    # STEPS_PER_EPOCH should be the number of instances \n    # divided by (GPU_COUNT*IMAGES_PER_GPU), and so should VALIDATION_STEPS;\n    # however, due to the time limit, I set them so that this kernel can be run in 9 hours\n    STEPS_PER_EPOCH = 1000\n    VALIDATION_STEPS = 200\n    \nconfig = FashionConfig()\nconfig.display()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:03:39.220952Z","iopub.execute_input":"2024-06-06T06:03:39.221350Z","iopub.status.idle":"2024-06-06T06:03:39.232622Z","shell.execute_reply.started":"2024-06-06T06:03:39.221278Z","shell.execute_reply":"2024-06-06T06:03:39.231792Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Make Datasets","metadata":{}},{"cell_type":"code","source":"with open(DATA_DIR/\"label_descriptions.json\") as f:\n    label_descriptions = json.load(f)\n\nlabel_names = [x['name'] for x in label_descriptions['categories']]","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:04:54.993354Z","iopub.execute_input":"2024-06-06T06:04:54.993650Z","iopub.status.idle":"2024-06-06T06:04:55.006400Z","shell.execute_reply.started":"2024-06-06T06:04:54.993607Z","shell.execute_reply":"2024-06-06T06:04:55.005537Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"segment_df = pd.read_csv(DATA_DIR/\"train.csv\")\n\nmultilabel_percent = len(segment_df[segment_df['ClassId'].str.contains('_')])/len(segment_df)*100\nprint(f\"Segments that have attributes: {multilabel_percent:.2f}%\")","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:05:21.181667Z","iopub.execute_input":"2024-06-06T06:05:21.182035Z","iopub.status.idle":"2024-06-06T06:05:48.534731Z","shell.execute_reply.started":"2024-06-06T06:05:21.181942Z","shell.execute_reply":"2024-06-06T06:05:48.534051Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Segments that contain attributes are only 3.46% of data, and [according to the host](https://www.kaggle.com/c/imaterialist-fashion-2019-FGVC6/discussion/90643#523135), 80% of images have no attribute. So, in the first step, we can only deal with categories to reduce the complexity of the task.","metadata":{}},{"cell_type":"code","source":"segment_df['CategoryId'] = segment_df['ClassId'].str.split('_').str[0]\n\nprint(\"Total segments: \", len(segment_df))\nsegment_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:06:14.781550Z","iopub.execute_input":"2024-06-06T06:06:14.781835Z","iopub.status.idle":"2024-06-06T06:06:15.709598Z","shell.execute_reply.started":"2024-06-06T06:06:14.781792Z","shell.execute_reply":"2024-06-06T06:06:15.708893Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Rows with the same image are grouped together because the subsequent operations perform in an image level.","metadata":{}},{"cell_type":"code","source":"image_df = segment_df.groupby('ImageId')['EncodedPixels', 'CategoryId'].agg(lambda x: list(x))\nsize_df = segment_df.groupby('ImageId')['Height', 'Width'].mean()\nimage_df = image_df.join(size_df, on='ImageId')\n\nprint(\"Total images: \", len(image_df))\nimage_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:06:51.034633Z","iopub.execute_input":"2024-06-06T06:06:51.035028Z","iopub.status.idle":"2024-06-06T06:07:00.706617Z","shell.execute_reply.started":"2024-06-06T06:06:51.034949Z","shell.execute_reply":"2024-06-06T06:07:00.705878Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Here is the custom function that resizes an image.","metadata":{}},{"cell_type":"code","source":"def resize_image(image_path):\n    img = cv2.imread(image_path)\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    img = cv2.resize(img, (IMAGE_SIZE, IMAGE_SIZE), interpolation=cv2.INTER_AREA)  \n    return img","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:07:22.210512Z","iopub.execute_input":"2024-06-06T06:07:22.210827Z","iopub.status.idle":"2024-06-06T06:07:22.215980Z","shell.execute_reply.started":"2024-06-06T06:07:22.210780Z","shell.execute_reply":"2024-06-06T06:07:22.215034Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"The crucial part is to create a dataset for this task.","metadata":{}},{"cell_type":"code","source":"class FashionDataset(utils.Dataset):\n\n    def __init__(self, df):\n        super().__init__(self)\n        \n        # Add classes\n        for i, name in enumerate(label_names):\n            self.add_class(\"fashion\", i+1, name)\n        \n        # Add images \n        for i, row in df.iterrows():\n            self.add_image(\"fashion\", \n                           image_id=row.name, \n                           path=str(DATA_DIR/'train'/row.name), \n                           labels=row['CategoryId'],\n                           annotations=row['EncodedPixels'], \n                           height=row['Height'], width=row['Width'])\n\n    def image_reference(self, image_id):\n        info = self.image_info[image_id]\n        return info['path'], [label_names[int(x)] for x in info['labels']]\n    \n    def load_image(self, image_id):\n        return resize_image(self.image_info[image_id]['path'])\n\n    def load_mask(self, image_id):\n        info = self.image_info[image_id]\n                \n        mask = np.zeros((IMAGE_SIZE, IMAGE_SIZE, len(info['annotations'])), dtype=np.uint8)\n        labels = []\n        \n        for m, (annotation, label) in enumerate(zip(info['annotations'], info['labels'])):\n            sub_mask = np.full(info['height']*info['width'], 0, dtype=np.uint8)\n            annotation = [int(x) for x in annotation.split(' ')]\n            \n            for i, start_pixel in enumerate(annotation[::2]):\n                sub_mask[start_pixel: start_pixel+annotation[2*i+1]] = 1\n\n            sub_mask = sub_mask.reshape((info['height'], info['width']), order='F')\n            sub_mask = cv2.resize(sub_mask, (IMAGE_SIZE, IMAGE_SIZE), interpolation=cv2.INTER_NEAREST)\n            \n            mask[:, :, m] = sub_mask\n            labels.append(int(label)+1)\n            \n        return mask, np.array(labels)","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:07:42.595851Z","iopub.execute_input":"2024-06-06T06:07:42.596147Z","iopub.status.idle":"2024-06-06T06:07:42.609077Z","shell.execute_reply.started":"2024-06-06T06:07:42.596103Z","shell.execute_reply":"2024-06-06T06:07:42.608091Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Let's visualize some random images and their masks.","metadata":{}},{"cell_type":"code","source":"dataset = FashionDataset(image_df)\ndataset.prepare()\n\nfor i in range(6):\n    image_id = random.choice(dataset.image_ids)\n    print(dataset.image_reference(image_id))\n    \n    image = dataset.load_image(image_id)\n    mask, class_ids = dataset.load_mask(image_id)\n    visualize.display_top_masks(image, mask, class_ids, dataset.class_names, limit=4)","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:07:59.179580Z","iopub.execute_input":"2024-06-06T06:07:59.179863Z","iopub.status.idle":"2024-06-06T06:08:10.683041Z","shell.execute_reply.started":"2024-06-06T06:07:59.179820Z","shell.execute_reply":"2024-06-06T06:08:10.682322Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Now, the data are partitioned into train and validation sets.","metadata":{}},{"cell_type":"code","source":"# This code partially supports k-fold training, \n# you can specify the fold to train and the total number of folds here\nFOLD = 0\nN_FOLDS = 5\n\nkf = KFold(n_splits=N_FOLDS, random_state=42, shuffle=True)\nsplits = kf.split(image_df) # ideally, this should be multilabel stratification\n\ndef get_fold():    \n    for i, (train_index, valid_index) in enumerate(splits):\n        if i == FOLD:\n            return image_df.iloc[train_index], image_df.iloc[valid_index]\n        \ntrain_df, valid_df = get_fold()\n\ntrain_dataset = FashionDataset(train_df)\ntrain_dataset.prepare()\n\nvalid_dataset = FashionDataset(valid_df)\nvalid_dataset.prepare()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:09:05.499462Z","iopub.execute_input":"2024-06-06T06:09:05.499814Z","iopub.status.idle":"2024-06-06T06:09:12.761691Z","shell.execute_reply.started":"2024-06-06T06:09:05.499752Z","shell.execute_reply":"2024-06-06T06:09:12.761023Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Let's visualize class distributions of the train and validation data.","metadata":{}},{"cell_type":"code","source":"train_segments = np.concatenate(train_df['CategoryId'].values).astype(int)\nprint(\"Total train images: \", len(train_df))\nprint(\"Total train segments: \", len(train_segments))\n\nplt.figure(figsize=(12, 3))\nvalues, counts = np.unique(train_segments, return_counts=True)\nplt.bar(values, counts)\nplt.xticks(values, label_names, rotation='vertical')\nplt.show()\n\nvalid_segments = np.concatenate(valid_df['CategoryId'].values).astype(int)\nprint(\"Total train images: \", len(valid_df))\nprint(\"Total validation segments: \", len(valid_segments))\n\nplt.figure(figsize=(12, 3))\nvalues, counts = np.unique(valid_segments, return_counts=True)\nplt.bar(values, counts)\nplt.xticks(values, label_names, rotation='vertical')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:09:57.622215Z","iopub.execute_input":"2024-06-06T06:09:57.622526Z","iopub.status.idle":"2024-06-06T06:09:59.452142Z","shell.execute_reply.started":"2024-06-06T06:09:57.622473Z","shell.execute_reply":"2024-06-06T06:09:59.451403Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Train","metadata":{}},{"cell_type":"code","source":"# Note that any hyperparameters here, such as LR, may still not be optimal\nLR = 1e-4\nEPOCHS = [2, 6, 8]\n\nimport warnings \nwarnings.filterwarnings(\"ignore\")","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:13:58.953313Z","iopub.execute_input":"2024-06-06T06:13:58.953648Z","iopub.status.idle":"2024-06-06T06:13:58.958759Z","shell.execute_reply.started":"2024-06-06T06:13:58.953600Z","shell.execute_reply":"2024-06-06T06:13:58.957725Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"This section creates a Mask R-CNN model and specifies augmentations to be used.","metadata":{}},{"cell_type":"code","source":"model = modellib.MaskRCNN(mode='training', config=config, model_dir=ROOT_DIR)\n\nmodel.load_weights(COCO_WEIGHTS_PATH, by_name=True, exclude=[\n    'mrcnn_class_logits', 'mrcnn_bbox_fc', 'mrcnn_bbox', 'mrcnn_mask'])","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:14:06.651909Z","iopub.execute_input":"2024-06-06T06:14:06.652263Z","iopub.status.idle":"2024-06-06T06:14:17.494042Z","shell.execute_reply.started":"2024-06-06T06:14:06.652212Z","shell.execute_reply":"2024-06-06T06:14:17.493103Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"augmentation = iaa.Sequential([\n    iaa.Fliplr(0.5),iaa.Sharpen(alpha=(0.0, 1.0), lightness=(0.75, 2.0)) # only horizontal flip here\n])","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:17:09.385475Z","iopub.execute_input":"2024-06-06T06:17:09.385819Z","iopub.status.idle":"2024-06-06T06:17:09.391968Z","shell.execute_reply.started":"2024-06-06T06:17:09.385771Z","shell.execute_reply":"2024-06-06T06:17:09.390906Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"First, we train only the heads.","metadata":{}},{"cell_type":"code","source":"%%time\nmodel.train(train_dataset, valid_dataset,\n            learning_rate=LR*2, # train heads with higher lr to speedup learning\n            epochs=EPOCHS[0],\n            layers='heads',\n            augmentation=None)\n\nhistory = model.keras_model.history.history","metadata":{"execution":{"iopub.status.busy":"2024-06-06T06:17:30.885714Z","iopub.execute_input":"2024-06-06T06:17:30.886024Z","iopub.status.idle":"2024-06-06T07:01:41.960030Z","shell.execute_reply.started":"2024-06-06T06:17:30.885966Z","shell.execute_reply":"2024-06-06T07:01:41.955540Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Then, all layers are trained.","metadata":{}},{"cell_type":"code","source":"%%time\nmodel.train(train_dataset, valid_dataset,\n            learning_rate=LR,\n            epochs=EPOCHS[1],\n            layers='all',\n            augmentation=augmentation)\n\nnew_history = model.keras_model.history.history\nfor k in new_history: history[k] = history[k] + new_history[k]","metadata":{"execution":{"iopub.status.busy":"2024-06-06T07:04:13.701497Z","iopub.execute_input":"2024-06-06T07:04:13.701878Z","iopub.status.idle":"2024-06-06T08:29:59.243392Z","shell.execute_reply.started":"2024-06-06T07:04:13.701810Z","shell.execute_reply":"2024-06-06T08:29:59.139619Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Afterwards, we reduce LR and train again.","metadata":{}},{"cell_type":"code","source":"%%time\nmodel.train(train_dataset, valid_dataset,\n            learning_rate=LR/5,\n            epochs=EPOCHS[2],\n            layers='all',\n            augmentation=augmentation)\n\nnew_history = model.keras_model.history.history\nfor k in new_history: history[k] = history[k] + new_history[k]","metadata":{"execution":{"iopub.status.busy":"2024-06-06T08:33:04.451256Z","iopub.execute_input":"2024-06-06T08:33:04.451567Z","iopub.status.idle":"2024-06-06T09:18:30.171070Z","shell.execute_reply.started":"2024-06-06T08:33:04.451521Z","shell.execute_reply":"2024-06-06T09:18:30.162491Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Let's visualize training history and choose the best epoch.","metadata":{}},{"cell_type":"code","source":"epochs = range(EPOCHS[-1])\n\nplt.figure(figsize=(18, 6))\n\nplt.subplot(131)\nplt.plot(epochs, history['loss'], label=\"train loss\")\nplt.plot(epochs, history['val_loss'], label=\"valid loss\")\nplt.legend()\nplt.subplot(132)\nplt.plot(epochs, history['mrcnn_class_loss'], label=\"train class loss\")\nplt.plot(epochs, history['val_mrcnn_class_loss'], label=\"valid class loss\")\nplt.legend()\nplt.subplot(133)\nplt.plot(epochs, history['mrcnn_mask_loss'], label=\"train mask loss\")\nplt.plot(epochs, history['val_mrcnn_mask_loss'], label=\"valid mask loss\")\nplt.legend()\n\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:32:04.434723Z","iopub.execute_input":"2024-06-06T09:32:04.435113Z","iopub.status.idle":"2024-06-06T09:32:05.333121Z","shell.execute_reply.started":"2024-06-06T09:32:04.435045Z","shell.execute_reply":"2024-06-06T09:32:05.332320Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"best_epoch = np.argmin(history[\"val_loss\"]) + 1\nprint(\"Best epoch: \", best_epoch)\nprint(\"Valid loss: \", history[\"val_loss\"][best_epoch-1])","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:32:34.777107Z","iopub.execute_input":"2024-06-06T09:32:34.777410Z","iopub.status.idle":"2024-06-06T09:32:34.783258Z","shell.execute_reply.started":"2024-06-06T09:32:34.777371Z","shell.execute_reply":"2024-06-06T09:32:34.782135Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Predict","metadata":{}},{"cell_type":"markdown","source":"The final step is to use our model to predict test data.","metadata":{}},{"cell_type":"code","source":"glob_list = glob.glob(f'/kaggle/working/fashion*/mask_rcnn_fashion_{best_epoch:04d}.h5')\nmodel_path = glob_list[0] if glob_list else ''","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:34:59.534089Z","iopub.execute_input":"2024-06-06T09:34:59.534398Z","iopub.status.idle":"2024-06-06T09:34:59.539615Z","shell.execute_reply.started":"2024-06-06T09:34:59.534355Z","shell.execute_reply":"2024-06-06T09:34:59.538815Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"This cell defines InferenceConfig and loads the best trained model.","metadata":{}},{"cell_type":"code","source":"class InferenceConfig(FashionConfig):\n    GPU_COUNT = 1\n    IMAGES_PER_GPU = 1\n\ninference_config = InferenceConfig()\n\nmodel = modellib.MaskRCNN(mode='inference', \n                          config=inference_config,\n                          model_dir=ROOT_DIR)\n\nassert model_path != '', \"Provide path to trained weights\"\nprint(\"Loading weights from \", model_path)\nmodel.load_weights(model_path, by_name=True)","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:35:06.287916Z","iopub.execute_input":"2024-06-06T09:35:06.288221Z","iopub.status.idle":"2024-06-06T09:35:17.397859Z","shell.execute_reply.started":"2024-06-06T09:35:06.288176Z","shell.execute_reply":"2024-06-06T09:35:17.397070Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Then, load the submission data.","metadata":{}},{"cell_type":"code","source":"sample_df = pd.read_csv(DATA_DIR/\"sample_submission.csv\")\nsample_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:36:32.648303Z","iopub.execute_input":"2024-06-06T09:36:32.648609Z","iopub.status.idle":"2024-06-06T09:36:32.686173Z","shell.execute_reply.started":"2024-06-06T09:36:32.648565Z","shell.execute_reply":"2024-06-06T09:36:32.685229Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Here is the main prediction steps, along with some helper functions.","metadata":{}},{"cell_type":"code","source":"# Convert data to run-length encoding\ndef to_rle(bits):\n    rle = []\n    pos = 0\n    for bit, group in itertools.groupby(bits):\n        group_list = list(group)\n        if bit:\n            rle.extend([pos, sum(group_list)])\n        pos += len(group_list)\n    return rle","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:37:10.356280Z","iopub.execute_input":"2024-06-06T09:37:10.356624Z","iopub.status.idle":"2024-06-06T09:37:10.361980Z","shell.execute_reply.started":"2024-06-06T09:37:10.356562Z","shell.execute_reply":"2024-06-06T09:37:10.361252Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Since the submission system does not permit overlapped masks, we have to fix them\ndef refine_masks(masks, rois):\n    areas = np.sum(masks.reshape(-1, masks.shape[-1]), axis=0)\n    mask_index = np.argsort(areas)\n    union_mask = np.zeros(masks.shape[:-1], dtype=bool)\n    for m in mask_index:\n        masks[:, :, m] = np.logical_and(masks[:, :, m], np.logical_not(union_mask))\n        union_mask = np.logical_or(masks[:, :, m], union_mask)\n    for m in range(masks.shape[-1]):\n        mask_pos = np.where(masks[:, :, m]==True)\n        if np.any(mask_pos):\n            y1, x1 = np.min(mask_pos, axis=1)\n            y2, x2 = np.max(mask_pos, axis=1)\n            rois[m, :] = [y1, x1, y2, x2]\n    return masks, rois","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:37:31.919903Z","iopub.execute_input":"2024-06-06T09:37:31.920236Z","iopub.status.idle":"2024-06-06T09:37:31.927799Z","shell.execute_reply.started":"2024-06-06T09:37:31.920177Z","shell.execute_reply":"2024-06-06T09:37:31.926662Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import cv2\nfrom tqdm import tqdm\n\ndef resize_image(image_path):\n    img = cv2.imread(image_path)\n    if img is None:\n        raise FileNotFoundError(f\"Image not found or unable to load: {image_path}\")\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    img = cv2.resize(img, (IMAGE_SIZE, IMAGE_SIZE), interpolation=cv2.INTER_AREA)\n    return img","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:46:03.882879Z","iopub.execute_input":"2024-06-06T09:46:03.883212Z","iopub.status.idle":"2024-06-06T09:46:03.888481Z","shell.execute_reply.started":"2024-06-06T09:46:03.883161Z","shell.execute_reply":"2024-06-06T09:46:03.887671Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\nsub_list = []\nmissing_count = 0\n\nfor i, row in tqdm(sample_df.iterrows(), total=len(sample_df)):\n    image_path = str(DATA_DIR/'test'/row['ImageId'])\n    try:\n        image = resize_image(image_path)\n        result = model.detect([image])[0]\n        \n        if result['masks'].size > 0:\n            masks, _ = refine_masks(result['masks'], result['rois'])\n            for m in range(masks.shape[-1]):\n                mask = masks[:, :, m].ravel(order='F')\n                rle = to_rle(mask)\n                label = result['class_ids'][m] - 1\n                sub_list.append([row['ImageId'], ' '.join(list(map(str, rle))), label])\n        else:\n            # 系统不允许缺少ID，这是填充它们的简单方法\n            sub_list.append([row['ImageId'], '1 1', 23])\n            missing_count += 1\n\n    except FileNotFoundError as e:\n        print(e)\n        # 如果图像未找到，添加一个默认条目\n        sub_list.append([row['ImageId'], '1 1', 23])\n        missing_count += 1\n    except Exception as e:\n        print(f\"Unexpected error occurred for image {row['ImageId']}: {e}\")\n        # 添加一个默认条目以处理其他异常情况\n        sub_list.append([row['ImageId'], '1 1', 23])\n        missing_count += 1\n\nprint(f\"Missing images count: {missing_count}\")","metadata":{"execution":{"iopub.status.busy":"2024-06-06T09:46:47.905597Z","iopub.execute_input":"2024-06-06T09:46:47.905884Z","iopub.status.idle":"2024-06-06T10:09:37.314776Z","shell.execute_reply.started":"2024-06-06T09:46:47.905843Z","shell.execute_reply":"2024-06-06T10:09:37.313594Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"The submission file is created, when all predictions are ready.","metadata":{}},{"cell_type":"code","source":"submission_df = pd.DataFrame(sub_list, columns=sample_df.columns.values)\nprint(\"Total image results: \", submission_df['ImageId'].nunique())\nprint(\"Missing Images: \", missing_count)\nsubmission_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T10:09:46.129503Z","iopub.execute_input":"2024-06-06T10:09:46.129790Z","iopub.status.idle":"2024-06-06T10:09:46.223690Z","shell.execute_reply.started":"2024-06-06T10:09:46.129748Z","shell.execute_reply":"2024-06-06T10:09:46.222976Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission_df.to_csv(\"submission.csv\", index=False)","metadata":{"execution":{"iopub.status.busy":"2024-06-06T10:10:09.218419Z","iopub.execute_input":"2024-06-06T10:10:09.218719Z","iopub.status.idle":"2024-06-06T10:10:09.989134Z","shell.execute_reply.started":"2024-06-06T10:10:09.218676Z","shell.execute_reply":"2024-06-06T10:10:09.988456Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Finally, it's pleasing to visualize the results! Sample images contain both fashion models and predictions from the Mask R-CNN model.","metadata":{}},{"cell_type":"code","source":"for i in range(9):\n    image_id = sample_df.sample()['ImageId'].values[0]\n    image_path = str(DATA_DIR/'test'/image_id)\n    \n    img = cv2.imread(image_path)\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    \n    result = model.detect([resize_image(image_path)])\n    r = result[0]\n    \n    if r['masks'].size > 0:\n        masks = np.zeros((img.shape[0], img.shape[1], r['masks'].shape[-1]), dtype=np.uint8)\n        for m in range(r['masks'].shape[-1]):\n            masks[:, :, m] = cv2.resize(r['masks'][:, :, m].astype('uint8'), \n                                        (img.shape[1], img.shape[0]), interpolation=cv2.INTER_NEAREST)\n        \n        y_scale = img.shape[0]/IMAGE_SIZE\n        x_scale = img.shape[1]/IMAGE_SIZE\n        rois = (r['rois'] * [y_scale, x_scale, y_scale, x_scale]).astype(int)\n        \n        masks, rois = refine_masks(masks, rois)\n    else:\n        masks, rois = r['masks'], r['rois']\n        \n    visualize.display_instances(img, rois, masks, r['class_ids'], \n                                ['bg']+label_names, r['scores'],\n                                title=image_id, figsize=(12, 12))","metadata":{"execution":{"iopub.status.busy":"2024-06-06T10:30:37.119041Z","iopub.execute_input":"2024-06-06T10:30:37.119399Z","iopub.status.idle":"2024-06-06T10:31:18.324975Z","shell.execute_reply.started":"2024-06-06T10:30:37.119340Z","shell.execute_reply":"2024-06-06T10:31:18.323712Z"},"collapsed":true,"jupyter":{"outputs_hidden":true},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample_df = pd.read_csv(\"/kaggle/input/varifying/varifying.csv\")\nsample_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T10:39:15.983149Z","iopub.execute_input":"2024-06-06T10:39:15.983454Z","iopub.status.idle":"2024-06-06T10:39:16.013141Z","shell.execute_reply.started":"2024-06-06T10:39:15.983410Z","shell.execute_reply":"2024-06-06T10:39:16.012241Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%time\nsub_list = []\nmissing_count = 0\n\nfor i, row in tqdm(sample_df.iterrows(), total=len(sample_df)):\n    image_path = os.path.join('/kaggle/input/varifying', row['ImageId'])\n    try:\n        image = resize_image(image_path)\n        result = model.detect([image])[0]\n        \n        if result['masks'].size > 0:\n            masks, _ = refine_masks(result['masks'], result['rois'])\n            for m in range(masks.shape[-1]):\n                mask = masks[:, :, m].ravel(order='F')\n                rle = to_rle(mask)\n                label = result['class_ids'][m] - 1\n                sub_list.append([row['ImageId'], ' '.join(list(map(str, rle))), label])\n        else:\n            # 系统不允许缺少ID，这是填充它们的简单方法\n            sub_list.append([row['ImageId'], '1 1', 23])\n            missing_count += 1\n\n    except FileNotFoundError as e:\n        print(e)\n        # 如果图像未找到，添加一个默认条目\n        sub_list.append([row['ImageId'], '1 1', 23])\n        missing_count += 1\n    except Exception as e:\n        print(f\"Unexpected error occurred for image {row['ImageId']}: {e}\")\n        # 添加一个默认条目以处理其他异常情况\n        sub_list.append([row['ImageId'], '1 1', 23])\n        missing_count += 1\n\nprint(f\"Missing images count: {missing_count}\")","metadata":{"execution":{"iopub.status.busy":"2024-06-06T10:48:25.542940Z","iopub.execute_input":"2024-06-06T10:48:25.543277Z","iopub.status.idle":"2024-06-06T10:48:28.514132Z","shell.execute_reply.started":"2024-06-06T10:48:25.543230Z","shell.execute_reply":"2024-06-06T10:48:28.513310Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission_df = pd.DataFrame(sub_list, columns=sample_df.columns.values)\nprint(\"Total image results: \", submission_df['ImageId'].nunique())\nprint(\"Missing Images: \", missing_count)\nsubmission_df.head()","metadata":{"execution":{"iopub.status.busy":"2024-06-06T11:03:02.194961Z","iopub.execute_input":"2024-06-06T11:03:02.195259Z","iopub.status.idle":"2024-06-06T11:03:02.214944Z","shell.execute_reply.started":"2024-06-06T11:03:02.195216Z","shell.execute_reply":"2024-06-06T11:03:02.213971Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"submission_df.to_csv(\"varifying.csv\", index=False)","metadata":{"execution":{"iopub.status.busy":"2024-06-06T11:03:07.138149Z","iopub.execute_input":"2024-06-06T11:03:07.138473Z","iopub.status.idle":"2024-06-06T11:03:07.145778Z","shell.execute_reply.started":"2024-06-06T11:03:07.138419Z","shell.execute_reply":"2024-06-06T11:03:07.144953Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for i in range(43):\n    image_id = sample_df.sample()['ImageId'].values[0]\n    image_path = os.path.join('/kaggle/input/varifying', image_id) \n    \n    img = cv2.imread(image_path)\n    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n    \n    result = model.detect([resize_image(image_path)])\n    r = result[0]\n    \n    if r['masks'].size > 0:\n        masks = np.zeros((img.shape[0], img.shape[1], r['masks'].shape[-1]), dtype=np.uint8)\n        for m in range(r['masks'].shape[-1]):\n            masks[:, :, m] = cv2.resize(r['masks'][:, :, m].astype('uint8'), \n                                        (img.shape[1], img.shape[0]), interpolation=cv2.INTER_NEAREST)\n        \n        y_scale = img.shape[0]/IMAGE_SIZE\n        x_scale = img.shape[1]/IMAGE_SIZE\n        rois = (r['rois'] * [y_scale, x_scale, y_scale, x_scale]).astype(int)\n        \n        masks, rois = refine_masks(masks, rois)\n    else:\n        masks, rois = r['masks'], r['rois']\n        \n    visualize.display_instances(img, rois, masks, r['class_ids'], \n                                ['bg']+label_names, r['scores'],\n                                title=image_id, figsize=(12, 12))","metadata":{"execution":{"iopub.status.busy":"2024-06-06T11:03:34.277458Z","iopub.execute_input":"2024-06-06T11:03:34.277772Z","iopub.status.idle":"2024-06-06T11:04:12.528690Z","shell.execute_reply.started":"2024-06-06T11:03:34.277724Z","shell.execute_reply":"2024-06-06T11:04:12.527974Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"My code is largely based on [this Mask-RCNN kernel](https://www.kaggle.com/hmendonca/mask-rcnn-and-coco-transfer-learning-lb-0-155) and borrowed some ideas from [the U-Net Baseline kernel](https://www.kaggle.com/go1dfish/u-net-baseline-by-pytorch-in-fgvc6-resize). So, I would like to thank the kernel authors for sharing insights and programming techniques. Importantly, an image segmentation task can be accomplished with short code and good accuracy thanks to [Matterport's implementation](https://github.com/matterport/Mask_RCNN) and a deep learning line of researches culminating in [Mask R-CNN](https://arxiv.org/abs/1703.06870).\n\nI am sorry that I published this kernel quite late, beyond the halfway of a timeline. I just started working for this competition about a week ago, and to my surprise, the score fell in the range of silver medals at that time. I have no dedicated GPU and no time to further tune the model, so I decided to make this kernel public as a starter guide for anyone who is interested to join this delightful competition.\n\n<img src='https://i.imgur.com/j6LPLQc.png'>","metadata":{}},{"cell_type":"markdown","source":"Hope you guys like this kernel. If there are any bugs, please let me know.\n\nP.S. When clicking 'Submit to Competition' button, I always run into 404 erros, so I have to save a submission file and upload it to the submission page for submitting. The public LB score of this kernel is around **0.07**.","metadata":{}}]}