Detectron2 with data preprocessing.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 json\nimport cv2\nfrom pathlib import Path\nimport matplotlib.pyplot as plt\nimport shutil\nimport random\nimport torch\nimport pandas as pd\nimport seaborn as sns","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:26.047820Z","iopub.execute_input":"2023-08-03T18:52:26.048689Z","iopub.status.idle":"2023-08-03T18:52:32.473251Z","shell.execute_reply.started":"2023-08-03T18:52:26.048641Z","shell.execute_reply":"2023-08-03T18:52:32.472115Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"COCO_PATH = Path(\"/kaggle/input/dlsprint2/badlad/labels/coco_format/train/badlad-train-coco.json\")\nwith open(COCO_PATH, \"r\") as f:\n    coco_data = json.load(f)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:32.478960Z","iopub.execute_input":"2023-08-03T18:52:32.481808Z","iopub.status.idle":"2023-08-03T18:52:39.577211Z","shell.execute_reply.started":"2023-08-03T18:52:32.481770Z","shell.execute_reply":"2023-08-03T18:52:39.576085Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"td=pd.DataFrame(coco_data['images'])\nprint(len(td))\ntd=pd.DataFrame(coco_data['annotations'])\nprint(len(td))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:39.581037Z","iopub.execute_input":"2023-08-03T18:52:39.581424Z","iopub.status.idle":"2023-08-03T18:52:41.203332Z","shell.execute_reply.started":"2023-08-03T18:52:39.581394Z","shell.execute_reply":"2023-08-03T18:52:41.202260Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#image Aspect Ratio\ntrain_images = pd.DataFrame( coco_data[\"images\"])\n\ntrain_images['aspect_ratio']=train_images['width']/train_images['height']\n\ntrain_images.rename(columns={\"id\":\"image_id\"}, inplace=True)\n\ntrain_images_aspect_ratio_count = train_images.groupby('aspect_ratio', as_index=False)['image_id'].nunique()\ntrain_images_aspect_ratio_count.rename(columns={'image_id':'image_count'}, inplace=True)\ntrain_images_aspect_ratio_count.sort_values(by='image_count', ascending=False,inplace=True)\n\nplt.figure(figsize=(15, 6))\nplt.title('Area vs Image Count Line Plot')\nsns.lineplot(x=train_images_aspect_ratio_count['aspect_ratio'], y = train_images_aspect_ratio_count['image_count'])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:41.206269Z","iopub.execute_input":"2023-08-03T18:52:41.207242Z","iopub.status.idle":"2023-08-03T18:52:41.808363Z","shell.execute_reply.started":"2023-08-03T18:52:41.207202Z","shell.execute_reply":"2023-08-03T18:52:41.807252Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#bbox Aspect Ratio\ntrain_annotations = pd.DataFrame(coco_data['annotations'])\n\ntrain_annotations.rename(columns={\"id\":\"annotation_id\"}, inplace=True)\nbbox_aspect_ratio=[]\nfor idx in train_annotations.index:\n    bbox_aspect_ratio.append(train_annotations['bbox'][idx][3]/train_annotations['bbox'][idx][2])\ntrain_annotations['bbox_aspect_ratio']=bbox_aspect_ratio\n\n\nbbox_aspect_ratio_count = train_annotations.groupby('bbox_aspect_ratio', as_index=False)['annotation_id'].count()\nbbox_aspect_ratio_count.rename(columns={'annotation_id':'annotation_count'}, inplace=True)\n\nplt.figure(figsize=(15, 6))\nplt.title('bbox aspect ratio vs Annotation Count Line Plot')\nsns.lineplot(x=bbox_aspect_ratio_count['bbox_aspect_ratio'], y = bbox_aspect_ratio_count['annotation_count'])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:41.810049Z","iopub.execute_input":"2023-08-03T18:52:41.811219Z","iopub.status.idle":"2023-08-03T18:52:47.188847Z","shell.execute_reply.started":"2023-08-03T18:52:41.811178Z","shell.execute_reply":"2023-08-03T18:52:47.180375Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n\n# low_aspect_ratio_threshold = 0.1\nhigh_aspect_ratio_threshold = 2.5\nhigh_bbox_aspect_ratio_threshold= 15.0","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.189815Z","iopub.status.idle":"2023-08-03T18:52:47.190239Z","shell.execute_reply.started":"2023-08-03T18:52:47.190032Z","shell.execute_reply":"2023-08-03T18:52:47.190052Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"images_to_keep = []\nimages_to_remove = []\nanno_to_keep = []\nanno_to_remove = []","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.192720Z","iopub.status.idle":"2023-08-03T18:52:47.193275Z","shell.execute_reply.started":"2023-08-03T18:52:47.192999Z","shell.execute_reply":"2023-08-03T18:52:47.193026Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for image_info in coco_data[\"images\"]:\n    image_id = image_info[\"id\"]\n    image_aspect_count = image_info[\"width\"] / image_info[\"height\"]\n    \n    image_aspect_count_tensor = torch.tensor(image_aspect_count, dtype=torch.float32, device=device)\n\n    if (image_aspect_count_tensor > high_aspect_ratio_threshold):\n        images_to_remove.append(image_id)\n    else:\n        images_to_keep.append(image_id)\n","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.195285Z","iopub.status.idle":"2023-08-03T18:52:47.195856Z","shell.execute_reply.started":"2023-08-03T18:52:47.195572Z","shell.execute_reply":"2023-08-03T18:52:47.195596Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for anno in coco_data[\"annotations\"]:\n    anno_id = anno['id']\n    bbox_aspect_ratio = anno['bbox'][3]/ anno['bbox'][2]\n    bbox_aspect_ratio_tensor = torch.tensor( bbox_aspect_ratio, dtype=torch.float32, device=device)\n    if bbox_aspect_ratio_tensor > high_bbox_aspect_ratio_threshold:\n        anno_to_remove.append(anno_id)\n    else:\n        anno_to_keep.append(anno_id)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.198170Z","iopub.status.idle":"2023-08-03T18:52:47.199428Z","shell.execute_reply.started":"2023-08-03T18:52:47.199126Z","shell.execute_reply":"2023-08-03T18:52:47.199155Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(len(anno_to_remove))\n# print(anno_to_remove[0:100])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.200863Z","iopub.status.idle":"2023-08-03T18:52:47.201875Z","shell.execute_reply.started":"2023-08-03T18:52:47.201573Z","shell.execute_reply":"2023-08-03T18:52:47.201600Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"image_category_counts = {}\nfor image_id in images_to_keep:\n        image_category_counts[image_id] = [0, 0, 0, 0]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.203247Z","iopub.status.idle":"2023-08-03T18:52:47.203996Z","shell.execute_reply.started":"2023-08-03T18:52:47.203714Z","shell.execute_reply":"2023-08-03T18:52:47.203738Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"for annotation in coco_data['annotations']:\n    image_id = annotation['image_id']\n    category_id = annotation['category_id']\n    if image_id in images_to_keep:\n        image_category_counts[image_id][category_id] += 1","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.205826Z","iopub.status.idle":"2023-08-03T18:52:47.206298Z","shell.execute_reply.started":"2023-08-03T18:52:47.206058Z","shell.execute_reply":"2023-08-03T18:52:47.206081Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"df_image_category_counts = pd.DataFrame.from_dict(image_category_counts, orient='index', columns=['Paragraph', 'Text-Box', 'Image', 'Table'])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.207640Z","iopub.status.idle":"2023-08-03T18:52:47.209085Z","shell.execute_reply.started":"2023-08-03T18:52:47.208836Z","shell.execute_reply":"2023-08-03T18:52:47.208860Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(df_image_category_counts)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.210212Z","iopub.status.idle":"2023-08-03T18:52:47.211330Z","shell.execute_reply.started":"2023-08-03T18:52:47.211021Z","shell.execute_reply":"2023-08-03T18:52:47.211056Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"filtered_images = df_image_category_counts[\n    (df_image_category_counts['Paragraph'] != 0) & (df_image_category_counts['Text-Box'] !=0) & (df_image_category_counts['Image'] ==0) & (df_image_category_counts['Table'] ==0)\n]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.212874Z","iopub.status.idle":"2023-08-03T18:52:47.213363Z","shell.execute_reply.started":"2023-08-03T18:52:47.213104Z","shell.execute_reply":"2023-08-03T18:52:47.213127Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(filtered_images)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.214998Z","iopub.status.idle":"2023-08-03T18:52:47.216719Z","shell.execute_reply.started":"2023-08-03T18:52:47.216446Z","shell.execute_reply":"2023-08-03T18:52:47.216473Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"filtered_images_final = filtered_images[\n    ((filtered_images['Paragraph'] > 100) | (filtered_images['Text-Box'] > 200))\n]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.218231Z","iopub.status.idle":"2023-08-03T18:52:47.219135Z","shell.execute_reply.started":"2023-08-03T18:52:47.218861Z","shell.execute_reply":"2023-08-03T18:52:47.218887Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(filtered_images_final)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.220751Z","iopub.status.idle":"2023-08-03T18:52:47.221667Z","shell.execute_reply.started":"2023-08-03T18:52:47.221394Z","shell.execute_reply":"2023-08-03T18:52:47.221420Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(len(filtered_images))\n# print(len(filtered_images_final))\n# print(list(filtered_images_final.index))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.223166Z","iopub.status.idle":"2023-08-03T18:52:47.223977Z","shell.execute_reply.started":"2023-08-03T18:52:47.223738Z","shell.execute_reply":"2023-08-03T18:52:47.223761Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(len(images_to_remove))\n# print(images_to_remove[0:100])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.225433Z","iopub.status.idle":"2023-08-03T18:52:47.226246Z","shell.execute_reply.started":"2023-08-03T18:52:47.225968Z","shell.execute_reply":"2023-08-03T18:52:47.225992Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"image_ids_to_delete = list(filtered_images_final.index)\nimage_ids_to_delete.extend(images_to_remove)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.227878Z","iopub.status.idle":"2023-08-03T18:52:47.228803Z","shell.execute_reply.started":"2023-08-03T18:52:47.228532Z","shell.execute_reply":"2023-08-03T18:52:47.228558Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(len(image_ids_to_delete))\n# print(image_ids_to_delete[0:100])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.230336Z","iopub.status.idle":"2023-08-03T18:52:47.231233Z","shell.execute_reply.started":"2023-08-03T18:52:47.230950Z","shell.execute_reply":"2023-08-03T18:52:47.230977Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"image_ids_to_delete=set(image_ids_to_delete)\nimage_ids_to_delete=list(image_ids_to_delete)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.232772Z","iopub.status.idle":"2023-08-03T18:52:47.233679Z","shell.execute_reply.started":"2023-08-03T18:52:47.233439Z","shell.execute_reply":"2023-08-03T18:52:47.233462Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"image_ids_to_delete=sorted(image_ids_to_delete)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.235120Z","iopub.status.idle":"2023-08-03T18:52:47.235974Z","shell.execute_reply.started":"2023-08-03T18:52:47.235715Z","shell.execute_reply":"2023-08-03T18:52:47.235740Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# print(len(image_ids_to_delete))\n# print(image_ids_to_delete[0:100])","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.237571Z","iopub.status.idle":"2023-08-03T18:52:47.238437Z","shell.execute_reply.started":"2023-08-03T18:52:47.238167Z","shell.execute_reply":"2023-08-03T18:52:47.238190Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"coco_data['images'] = [image_info for image_info in coco_data['images'] if image_info['id'] not in image_ids_to_delete]\ncoco_data['annotations'] = [annotation for annotation in coco_data['annotations'] if (annotation['image_id'] not in image_ids_to_delete)]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.239904Z","iopub.status.idle":"2023-08-03T18:52:47.240769Z","shell.execute_reply.started":"2023-08-03T18:52:47.240476Z","shell.execute_reply":"2023-08-03T18:52:47.240507Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"coco_data['annotations'] = [annotation for annotation in coco_data['annotations'] if (annotation['id'] not in anno_to_remove)]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.242372Z","iopub.status.idle":"2023-08-03T18:52:47.243189Z","shell.execute_reply.started":"2023-08-03T18:52:47.242929Z","shell.execute_reply":"2023-08-03T18:52:47.242954Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"td=pd.DataFrame(coco_data['annotations'])\nprint(\"-----Train Anotation: \",len(td))\ntd=pd.DataFrame(coco_data['images'])\nprint(\"-----Train Image: \",len(td))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.244709Z","iopub.status.idle":"2023-08-03T18:52:47.245507Z","shell.execute_reply.started":"2023-08-03T18:52:47.245237Z","shell.execute_reply":"2023-08-03T18:52:47.245270Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"filtered_json_file_path = 'filtered_badlad-train-coco.json'\nwith open(filtered_json_file_path, 'w') as filtered_json_file:\n    json.dump(coco_data, filtered_json_file)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.247097Z","iopub.status.idle":"2023-08-03T18:52:47.248273Z","shell.execute_reply.started":"2023-08-03T18:52:47.247987Z","shell.execute_reply":"2023-08-03T18:52:47.248014Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"kaggle_working_directory = '/kaggle/working/'\nshutil.move(filtered_json_file_path, Path(kaggle_working_directory) / filtered_json_file_path)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.249984Z","iopub.status.idle":"2023-08-03T18:52:47.250869Z","shell.execute_reply.started":"2023-08-03T18:52:47.250575Z","shell.execute_reply":"2023-08-03T18:52:47.250620Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(\"Preprocessing Done\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.252439Z","iopub.status.idle":"2023-08-03T18:52:47.253240Z","shell.execute_reply.started":"2023-08-03T18:52:47.253002Z","shell.execute_reply":"2023-08-03T18:52:47.253025Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"%%capture\nimport sys, os, distutils.core\n!git clone 'https://github.com/facebookresearch/detectron2'\ndist = distutils.core.run_setup(\"./detectron2/setup.py\")\n!python -m pip install {' '.join([f\"'{x}'\" for x in dist.install_requires])}\nsys.path.insert(0, os.path.abspath('./detectron2'))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.254726Z","iopub.status.idle":"2023-08-03T18:52:47.255591Z","shell.execute_reply.started":"2023-08-03T18:52:47.255328Z","shell.execute_reply":"2023-08-03T18:52:47.255354Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from datetime import datetime\nis_train = True\nis_evaluate = False\nis_inference = True\nis_resume_training = False\nis_augment = False\nSEED = 1081983","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.257008Z","iopub.status.idle":"2023-08-03T18:52:47.257842Z","shell.execute_reply.started":"2023-08-03T18:52:47.257577Z","shell.execute_reply":"2023-08-03T18:52:47.257602Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from pathlib import Path\n\nTRAIN_IMG_DIR = Path(\"/kaggle/input/dlsprint2/badlad/images/train\")\nTRAIN_COCO_PATH = Path(\"/kaggle/working/filtered_badlad-train-coco.json\")\n\nTEST_IMG_DIR = Path(\"/kaggle/input/dlsprint2/badlad/images/test\")\nTEST_METADATA_PATH = Path(\"/kaggle/input/dlsprint2/badlad/badlad-test-metadata.json\")\n\nOUTPUT_DIR = Path(\"./output\")\nOUTPUT_MODEL = OUTPUT_DIR/\"model_final.pth\"\n\nPRETRAINED_PATH = Path(\"\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.259283Z","iopub.status.idle":"2023-08-03T18:52:47.260120Z","shell.execute_reply.started":"2023-08-03T18:52:47.259865Z","shell.execute_reply":"2023-08-03T18:52:47.259890Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"MODEL_PATH = OUTPUT_MODEL if is_train else PRETRAINED_PATH","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.261700Z","iopub.status.idle":"2023-08-03T18:52:47.262541Z","shell.execute_reply.started":"2023-08-03T18:52:47.262257Z","shell.execute_reply":"2023-08-03T18:52:47.262290Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"from detectron2.utils.memory import retry_if_cuda_oom\nfrom detectron2.utils.logger import setup_logger\nfrom detectron2.checkpoint import DetectionCheckpointer\nfrom detectron2.modeling import build_model\nfrom detectron2.evaluation import COCOEvaluator, inference_on_dataset\nimport detectron2.data.transforms as T\nfrom detectron2.data import detection_utils as utils\nfrom detectron2.data import DatasetCatalog, MetadataCatalog, build_detection_test_loader, build_detection_train_loader, DatasetMapper\nfrom detectron2.utils.visualizer import Visualizer\nfrom detectron2.structures import BoxMode\nfrom detectron2.engine import DefaultPredictor, DefaultTrainer\nfrom detectron2.config import get_cfg\nfrom detectron2 import model_zoo\n\nimport pandas as pd\nimport numpy as np\nfrom tqdm.notebook import tqdm \nimport matplotlib.pyplot as plt\nimport json\nimport cv2\nimport copy\nfrom typing import Optional\n\nfrom IPython.display import FileLink\nimport torch\nimport gc\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\nsetup_logger()","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.264015Z","iopub.status.idle":"2023-08-03T18:52:47.264847Z","shell.execute_reply.started":"2023-08-03T18:52:47.264607Z","shell.execute_reply":"2023-08-03T18:52:47.264630Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with TRAIN_COCO_PATH.open() as f:\n    train_dict = json.load(f)\n\nwith TEST_METADATA_PATH.open() as f:\n    test_dict = json.load(f)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.266219Z","iopub.status.idle":"2023-08-03T18:52:47.267057Z","shell.execute_reply.started":"2023-08-03T18:52:47.266800Z","shell.execute_reply":"2023-08-03T18:52:47.266824Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def organize_coco_data(data_dict: dict) -> tuple[list[str], list[dict], list[dict]]:\n    thing_classes: list[str] = []\n\n    for cat in data_dict['categories']:\n        thing_classes.append(cat['name'])\n\n    images_metadata: list[dict] = data_dict['images']\n    data_annotations = []\n    for ann in data_dict['annotations']:\n        annot_obj = {\n            \"id\": ann['id'],\n            \"gt_masks\": ann['segmentation'],\n            \"image_id\": ann['image_id'],\n            \"category_id\": ann['category_id'],\n\n            \"x_min\": ann['bbox'][0], \n            \"y_min\": ann['bbox'][1],  \n            \"x_max\": ann['bbox'][0] + ann['bbox'][2], \n            \"y_max\": ann['bbox'][1] + ann['bbox'][3] \n        }\n        data_annotations.append(annot_obj)\n    return thing_classes, images_metadata, data_annotations","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.268571Z","iopub.status.idle":"2023-08-03T18:52:47.269381Z","shell.execute_reply.started":"2023-08-03T18:52:47.269110Z","shell.execute_reply":"2023-08-03T18:52:47.269133Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"thing_classes, images_metadata, data_annotations = organize_coco_data(\n    train_dict\n)\n\nthing_classes_test, images_metadata_test, _ = organize_coco_data(\n    test_dict\n)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.270918Z","iopub.status.idle":"2023-08-03T18:52:47.271846Z","shell.execute_reply.started":"2023-08-03T18:52:47.271554Z","shell.execute_reply":"2023-08-03T18:52:47.271581Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(thing_classes)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.273287Z","iopub.status.idle":"2023-08-03T18:52:47.274130Z","shell.execute_reply.started":"2023-08-03T18:52:47.273878Z","shell.execute_reply":"2023-08-03T18:52:47.273902Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_metadata = pd.DataFrame(images_metadata)\ntrain_metadata = train_metadata[['id', 'file_name', 'width', 'height']]\ntrain_metadata = train_metadata.rename(columns={\"id\": \"image_id\"})\nprint(\"train_metadata size=\", len(train_metadata))\ntrain_metadata.head(5)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.275532Z","iopub.status.idle":"2023-08-03T18:52:47.276393Z","shell.execute_reply.started":"2023-08-03T18:52:47.276112Z","shell.execute_reply":"2023-08-03T18:52:47.276136Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_annot_df = pd.DataFrame(data_annotations)\nprint(\"train_annot_df size=\", len(train_annot_df))\ntrain_annot_df.head(5)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.277871Z","iopub.status.idle":"2023-08-03T18:52:47.278655Z","shell.execute_reply.started":"2023-08-03T18:52:47.278415Z","shell.execute_reply":"2023-08-03T18:52:47.278438Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test_metadata = pd.DataFrame(images_metadata_test)\ntest_metadata = test_metadata[['id', 'file_name', 'width', 'height']]\ntest_metadata = test_metadata.rename(columns={\"id\": \"image_id\"})\nprint(\"test_metadata size=\", len(test_metadata))\ntest_metadata.head(5)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.280088Z","iopub.status.idle":"2023-08-03T18:52:47.281004Z","shell.execute_reply.started":"2023-08-03T18:52:47.280707Z","shell.execute_reply":"2023-08-03T18:52:47.280734Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"TRAIN_SPLIT = 0.90","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.286951Z","iopub.status.idle":"2023-08-03T18:52:47.287842Z","shell.execute_reply.started":"2023-08-03T18:52:47.287535Z","shell.execute_reply":"2023-08-03T18:52:47.287560Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"n_dataset = len(train_metadata)\nn_train = int(n_dataset * TRAIN_SPLIT)\nprint(\"n_dataset\", n_dataset, \"n_train\", n_train, \"n_val\", n_dataset-n_train)\n\nnp.random.seed(SEED)\n\ninds = np.random.permutation(n_dataset)\ntrain_inds, valid_inds = inds[:n_train], inds[n_train:]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.289383Z","iopub.status.idle":"2023-08-03T18:52:47.290160Z","shell.execute_reply.started":"2023-08-03T18:52:47.289918Z","shell.execute_reply":"2023-08-03T18:52:47.289942Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def convert_coco_to_detectron2_format(\n    imgdir: Path,\n    metadata_df: pd.DataFrame,\n    annot_df: Optional[pd.DataFrame] = None,\n    target_indices: Optional[np.ndarray] = None,\n):\n\n    dataset_dicts = []\n    for _, train_meta_row in tqdm(metadata_df.iterrows(), total=len(metadata_df)):\n        image_id, filename, width, height = train_meta_row.values\n\n        annotations = []\n        if annot_df is not None:\n            for _, ann in annot_df.query(\"image_id == @image_id\").iterrows():\n                class_id = ann[\"category_id\"]\n                gt_masks = ann[\"gt_masks\"]\n                bbox_resized = [\n                    float(ann[\"x_min\"]),\n                    float(ann[\"y_min\"]),\n                    float(ann[\"x_max\"]),\n                    float(ann[\"y_max\"]),\n                ]\n\n                annotation = {\n                    \"bbox\": bbox_resized,\n                    \"bbox_mode\": BoxMode.XYXY_ABS,\n                    \"segmentation\": gt_masks,\n                    \"category_id\": class_id,\n                }\n\n                annotations.append(annotation)\n        record = {\n            \"file_name\": str(imgdir/filename),\n            \"image_id\": image_id,\n            \"width\": width,\n            \"height\": height,\n            \"annotations\": annotations\n        }\n\n        dataset_dicts.append(record)\n\n    if target_indices is not None:\n        dataset_dicts = [dataset_dicts[i] for i in target_indices]\n\n    return dataset_dicts","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.291769Z","iopub.status.idle":"2023-08-03T18:52:47.292659Z","shell.execute_reply.started":"2023-08-03T18:52:47.292353Z","shell.execute_reply":"2023-08-03T18:52:47.292378Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"DATA_REGISTER_TRAINING = \"badlad_train\"\nDATA_REGISTER_VALID    = \"badlad_valid\"\nDATA_REGISTER_TEST     = \"badlad_test\"","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.294162Z","iopub.status.idle":"2023-08-03T18:52:47.295006Z","shell.execute_reply.started":"2023-08-03T18:52:47.294753Z","shell.execute_reply":"2023-08-03T18:52:47.294778Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_train:\n    DatasetCatalog.register(\n        DATA_REGISTER_TRAINING,\n        lambda: convert_coco_to_detectron2_format(\n            TRAIN_IMG_DIR,\n            train_metadata,\n            train_annot_df,\n            target_indices=train_inds,\n        ),\n    )\n\n    MetadataCatalog.get(DATA_REGISTER_TRAINING).set(thing_classes=thing_classes)\n\n    dataset_dicts_train = DatasetCatalog.get(DATA_REGISTER_TRAINING)\n    metadata_dicts_train = MetadataCatalog.get(DATA_REGISTER_TRAINING)\n\n    print(\"dicts training size=\", len(dataset_dicts_train))\n    print(\"-----------------------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.296501Z","iopub.status.idle":"2023-08-03T18:52:47.297342Z","shell.execute_reply.started":"2023-08-03T18:52:47.297057Z","shell.execute_reply":"2023-08-03T18:52:47.297080Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_train or is_evaluate:\n    DatasetCatalog.register(\n        DATA_REGISTER_VALID,\n        lambda: convert_coco_to_detectron2_format(\n            TRAIN_IMG_DIR,\n            train_metadata,\n            train_annot_df,\n            target_indices=valid_inds,\n        ),\n    )\n    \n    MetadataCatalog.get(DATA_REGISTER_VALID).set(thing_classes=thing_classes)\n\n    dataset_dicts_valid = DatasetCatalog.get(DATA_REGISTER_VALID)\n    metadata_dicts_valid = MetadataCatalog.get(DATA_REGISTER_VALID)\n\n    print(\"dicts valid size=\", len(dataset_dicts_valid))\n    print(\"-------------------------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.298817Z","iopub.status.idle":"2023-08-03T18:52:47.299638Z","shell.execute_reply.started":"2023-08-03T18:52:47.299385Z","shell.execute_reply":"2023-08-03T18:52:47.299409Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"DatasetCatalog.register(\n    DATA_REGISTER_TEST,\n    lambda: convert_coco_to_detectron2_format(\n        TEST_IMG_DIR,\n        test_metadata,\n    )\n)\n\nMetadataCatalog.get(DATA_REGISTER_TEST).set(\n    thing_classes=thing_classes_test\n)\n\ndataset_dicts_test = DatasetCatalog.get(DATA_REGISTER_TEST)\nmetadata_dicts_test = MetadataCatalog.get(DATA_REGISTER_TEST)\nprint(\"dicts test size=\", len(dataset_dicts_test))\nprint(\"-------------------------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.301174Z","iopub.status.idle":"2023-08-03T18:52:47.302081Z","shell.execute_reply.started":"2023-08-03T18:52:47.301844Z","shell.execute_reply":"2023-08-03T18:52:47.301868Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(\"---------Data is Ready-----------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.303669Z","iopub.status.idle":"2023-08-03T18:52:47.304532Z","shell.execute_reply.started":"2023-08-03T18:52:47.304232Z","shell.execute_reply":"2023-08-03T18:52:47.304258Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def custom_mapper(dataset_dict):\n    dataset_dict = copy.deepcopy(dataset_dict)\n    image = utils.read_image(dataset_dict[\"file_name\"], format=\"BGR\")\n\n    transform_list = [\n                        T.RandomRotation(angle=[-7, 7]),\n#                         T.RandomScale(scale=(0.2, 1.2)),\n                        T.RandomFlip(prob=0.5, horizontal=False, vertical=True),\n                        T.RandomFlip(prob=0.5, horizontal=True, vertical=False),\n#                         T.RandomBrightness(0.5, 1.2),\n#                         T.RandomContrast(0.5, 1.2),\n#                         T.RandomSaturation(0.5, 1.2),\n#                         T.RandomHue(0.1)\n                    ]    \n                    \n    image, transforms = T.apply_transform_gens(transform_list, image)\n\n    dataset_dict[\"image\"] = torch.as_tensor(\n        image.transpose(2, 0, 1).astype(\"float32\"))\n\n    annos = [\n        utils.transform_instance_annotations(obj, transforms, image.shape[:2])\n        for obj in dataset_dict.pop(\"annotations\")\n        if obj.get(\"iscrowd\", 0) == 0\n    ]\n    instances = utils.annotations_to_instances(annos, image.shape[:2])\n\n    dataset_dict[\"instances\"] = utils.filter_empty_instances(instances)\n\n    return dataset_dict","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.306000Z","iopub.status.idle":"2023-08-03T18:52:47.306838Z","shell.execute_reply.started":"2023-08-03T18:52:47.306582Z","shell.execute_reply":"2023-08-03T18:52:47.306606Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class AugTrainer(DefaultTrainer):\n    @classmethod\n    def build_train_loader(cls, cfg):\n        return build_detection_train_loader(cfg, mapper=custom_mapper)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.308381Z","iopub.status.idle":"2023-08-03T18:52:47.309224Z","shell.execute_reply.started":"2023-08-03T18:52:47.308968Z","shell.execute_reply":"2023-08-03T18:52:47.308993Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_train:\n    cfg = get_cfg()\n    \n    config_name = \"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml\"\n\n    cfg.merge_from_file(model_zoo.get_config_file(config_name))\n\n    cfg.DATASETS.TRAIN = (DATA_REGISTER_TRAINING,)\n    cfg.DATASETS.TEST = (DATA_REGISTER_VALID,)\n\n    cfg.DATALOADER.NUM_WORKERS = 2\n\n    if (is_resume_training):\n        print(\"#### SETTING PRETRAINED WEIGHTS TO RESUME TRAINING ####\")\n        cfg.MODEL.WEIGHTS = str(PRETRAINED_PATH)\n    else:\n        print(\"#### TRAINING MODEL FROM SCRATCH ####\")\n\n    cfg.SOLVER.AMP.ENABLED = True\n    cfg.SOLVER.IMS_PER_BATCH = 8\n    cfg.SOLVER.BASE_LR = 0.01\n    cfg.SOLVER.WARMUP_ITERS = 20\n    cfg.SOLVER.MAX_ITER = 18000\n    cfg.SOLVER.GAMMA = 0.005\n    cfg.SOLVER.CHECKPOINT_PERIOD = 500\n    cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 180\n    cfg.MODEL.ROI_HEADS.NUM_CLASSES = 4\n\n    cfg.OUTPUT_DIR = str(OUTPUT_DIR)\n    print(\"creating cfg.OUTPUT_DIR -> \", cfg.OUTPUT_DIR)\n    OUTPUT_DIR.mkdir(exist_ok=True)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.310766Z","iopub.status.idle":"2023-08-03T18:52:47.311660Z","shell.execute_reply.started":"2023-08-03T18:52:47.311385Z","shell.execute_reply":"2023-08-03T18:52:47.311412Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_train:\n    trainer = DefaultTrainer(cfg) if not is_augment else AugTrainer(cfg)\n        \n    trainer.resume_or_load(resume=is_resume_training)\n\n    trainer.train()\n    \n    print(\"----Training Finish----\")\n    _ = trainer.model.train(False)\n    \n    FileLink(str(OUTPUT_MODEL))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.313179Z","iopub.status.idle":"2023-08-03T18:52:47.314049Z","shell.execute_reply.started":"2023-08-03T18:52:47.313793Z","shell.execute_reply":"2023-08-03T18:52:47.313818Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_train:\n    metrics_df = pd.read_json(\n        OUTPUT_DIR/\"metrics.json\", orient=\"records\", lines=True\n    )\n    mdf = metrics_df.sort_values(\"iteration\")\n    print(mdf.head(10).T)\n    fig, ax = plt.subplots()\n\n    mdf1 = mdf[~mdf[\"total_loss\"].isna()]\n    ax.plot(mdf1[\"iteration\"], mdf1[\"total_loss\"], c=\"C0\", label=\"train\")\n\n    if \"validation_loss\" in mdf.columns:\n        mdf2 = mdf[~mdf[\"validation_loss\"].isna()]\n        ax.plot(mdf2[\"iteration\"], mdf2[\"validation_loss\"],\n                c=\"C1\", label=\"validation\")\n\n    ax.legend()\n    ax.set_title(\"Loss curve\")\n    plt.show()\n    fig, ax = plt.subplots()\n\n    mdf1 = mdf[~mdf[\"fast_rcnn/cls_accuracy\"].isna()]\n    ax.plot(mdf1[\"iteration\"], mdf1[\"fast_rcnn/cls_accuracy\"],\n            c=\"C0\", label=\"train\")\n\n    ax.legend()\n    ax.set_title(\"Accuracy curve\")\n    plt.show()\n    fig, ax = plt.subplots()\n\n    mdf1 = mdf[~mdf[\"loss_box_reg\"].isna()]\n    ax.plot(mdf1[\"iteration\"], mdf1[\"loss_box_reg\"], c=\"C0\", label=\"train\")\n\n    ax.legend()\n    ax.set_title(\"loss_box_reg\")\n    plt.show()","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.315579Z","iopub.status.idle":"2023-08-03T18:52:47.316445Z","shell.execute_reply.started":"2023-08-03T18:52:47.316138Z","shell.execute_reply":"2023-08-03T18:52:47.316162Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_evaluate:\n    print(\"------Evaluating on Validation data-----------\")\n    cfg.MODEL.WEIGHTS = str(MODEL_PATH)\n    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5\n\n    cfg.SOLVER.IMS_PER_BATCH = 64\n\n    evaluator = COCOEvaluator(\n        DATA_REGISTER_VALID, cfg, False, output_dir=cfg.OUTPUT_DIR, use_fast_impl=True\n    )\n\n    val_loader = build_detection_test_loader(cfg, DATA_REGISTER_VALID)\n\n    results = inference_on_dataset(\n        trainer.model, val_loader, evaluator=evaluator\n    )","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.318022Z","iopub.status.idle":"2023-08-03T18:52:47.318934Z","shell.execute_reply.started":"2023-08-03T18:52:47.318661Z","shell.execute_reply":"2023-08-03T18:52:47.318687Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"inf_cfg = get_cfg()\n\nconfig_name = \"COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml\"\n\ninf_cfg.merge_from_file(model_zoo.get_config_file(config_name))\ninf_cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 180\ninf_cfg.MODEL.ROI_HEADS.NUM_CLASSES = 4\ninf_cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5\ninf_cfg.MODEL.DEVICE = \"cuda\"\n\ninf_cfg.DATALOADER.NUM_WORKERS = 2 \ninf_cfg.MODEL.WEIGHTS = str(MODEL_PATH)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.320535Z","iopub.status.idle":"2023-08-03T18:52:47.321476Z","shell.execute_reply.started":"2023-08-03T18:52:47.321161Z","shell.execute_reply":"2023-08-03T18:52:47.321188Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"BATCH = 10\ntest_loader = build_detection_test_loader(inf_cfg, DATA_REGISTER_TEST, batch_size=BATCH)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.323158Z","iopub.status.idle":"2023-08-03T18:52:47.323978Z","shell.execute_reply.started":"2023-08-03T18:52:47.323723Z","shell.execute_reply":"2023-08-03T18:52:47.323748Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"ACCEPTANCE_THRESHOLD = 0.6","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.325567Z","iopub.status.idle":"2023-08-03T18:52:47.326393Z","shell.execute_reply.started":"2023-08-03T18:52:47.326110Z","shell.execute_reply":"2023-08-03T18:52:47.326134Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(f\"------------ MODEL: {inf_cfg.MODEL.WEIGHTS} FOR INFERENCE --------------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.327964Z","iopub.status.idle":"2023-08-03T18:52:47.328865Z","shell.execute_reply.started":"2023-08-03T18:52:47.328595Z","shell.execute_reply":"2023-08-03T18:52:47.328621Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"predictor = DefaultPredictor(inf_cfg)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.331714Z","iopub.status.idle":"2023-08-03T18:52:47.332571Z","shell.execute_reply.started":"2023-08-03T18:52:47.332303Z","shell.execute_reply":"2023-08-03T18:52:47.332343Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"fig, ax = plt.subplots(2, 2, figsize=(20, 20))\nindices = [ax[0][0], ax[1][0], ax[0][1], ax[1][1]]\n\n# Show some qualitative results by predicting on test set images\nNUM_TEST_SAMPLES = 4\nsamples = np.random.choice(dataset_dicts_test, NUM_TEST_SAMPLES)\n\nfor i, sample in enumerate(samples):\n    img = cv2.imread(sample[\"file_name\"])\n    outputs = predictor(img)\n    visualizer = Visualizer(img, metadata=metadata_dicts_test, scale=0.5,)\n    visualizer = visualizer.draw_instance_predictions(\n        outputs[\"instances\"].to(\"cpu\")\n    )\n    display_img = visualizer.get_image()[:, :, ::-1]\n    indices[i].grid(False)\n    indices[i].imshow(display_img)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.334023Z","iopub.status.idle":"2023-08-03T18:52:47.334873Z","shell.execute_reply.started":"2023-08-03T18:52:47.334617Z","shell.execute_reply":"2023-08-03T18:52:47.334641Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def rebuild_model():\n    model = build_model(inf_cfg)\n    _ = DetectionCheckpointer(model).load(inf_cfg.MODEL.WEIGHTS)\n    return model","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.336372Z","iopub.status.idle":"2023-08-03T18:52:47.337210Z","shell.execute_reply.started":"2023-08-03T18:52:47.336950Z","shell.execute_reply":"2023-08-03T18:52:47.336974Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"model = rebuild_model()","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.338680Z","iopub.status.idle":"2023-08-03T18:52:47.339502Z","shell.execute_reply.started":"2023-08-03T18:52:47.339221Z","shell.execute_reply":"2023-08-03T18:52:47.339246Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!export LRU_CACHE_CAPACITY=1\n!export 'PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:512'","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.341040Z","iopub.status.idle":"2023-08-03T18:52:47.341953Z","shell.execute_reply.started":"2023-08-03T18:52:47.341672Z","shell.execute_reply":"2023-08-03T18:52:47.341699Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"vars_to_del = [\"trainer\", \"predictor\", \"outputs\"]\n\nfor v in vars_to_del:\n    if v in globals():\n        print(f\"Deleting {v}\")\n        del globals()[v]\n    elif v in locals():\n        print(f\"Deleting {v}\")\n        del locals()[v]","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.343417Z","iopub.status.idle":"2023-08-03T18:52:47.344185Z","shell.execute_reply.started":"2023-08-03T18:52:47.343942Z","shell.execute_reply":"2023-08-03T18:52:47.343966Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def rle_encode(mask):\n    pixels = mask.T.flatten()\n    use_padding = False\n    if pixels[0] or pixels[-1]:\n        use_padding = True\n        pixel_padded = np.zeros([len(pixels) + 2], dtype=pixels.dtype)\n        pixel_padded[1:-1] = pixels\n        pixels = pixel_padded\n    rle = np.where(pixels[1:] != pixels[:-1])[0] + 2\n    if use_padding:\n        rle = rle - 1\n    rle[1::2] = rle[1::2] - rle[:-1:2]\n    return ' '.join(str(x) for x in rle)","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.345752Z","iopub.status.idle":"2023-08-03T18:52:47.346550Z","shell.execute_reply.started":"2023-08-03T18:52:47.346277Z","shell.execute_reply":"2023-08-03T18:52:47.346299Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"@retry_if_cuda_oom\ndef get_masks(prediction):\n    # get masks for each category\n    take = prediction.scores >= ACCEPTANCE_THRESHOLD\n    pred_masks = (prediction.pred_masks[take] != 0)\n    pred_classes = prediction.pred_classes[take]\n  \n    rles = []\n    for cat in range(len(thing_classes)):\n        pred_mask = pred_masks[pred_classes == cat]\n        \n        # pred_mask = retry_if_cuda_oom(torch.any)(pred_mask, dim=0)\n        pred_mask = torch.any(pred_mask, dim=0)\n        rles.append(rle_encode(pred_mask.short().to(\"cpu\").numpy()))\n\n    return rles","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.348084Z","iopub.status.idle":"2023-08-03T18:52:47.348972Z","shell.execute_reply.started":"2023-08-03T18:52:47.348693Z","shell.execute_reply":"2023-08-03T18:52:47.348719Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def run_inference(data):\n    results = []\n    with torch.no_grad():\n        outputs = model(data)\n        if torch.cuda.is_available():\n            torch.cuda.synchronize()\n\n        for idx, output in enumerate(outputs):\n            output = output[\"instances\"]\n\n            rles = get_masks(output)\n\n            result = [\n                f\"{data[idx]['image_id']}_{cat},{rles[cat]}\\n\"\n                for cat in range(len(thing_classes))\n            ]\n\n            results.extend(result)\n\n        del outputs, output\n\n    return results","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.350472Z","iopub.status.idle":"2023-08-03T18:52:47.351409Z","shell.execute_reply.started":"2023-08-03T18:52:47.351090Z","shell.execute_reply":"2023-08-03T18:52:47.351117Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(\"---------Inferance Start----------\")","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.352960Z","iopub.status.idle":"2023-08-03T18:52:47.353770Z","shell.execute_reply.started":"2023-08-03T18:52:47.353526Z","shell.execute_reply":"2023-08-03T18:52:47.353549Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"torch.cuda.empty_cache()\ngc.collect()","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.355188Z","iopub.status.idle":"2023-08-03T18:52:47.356112Z","shell.execute_reply.started":"2023-08-03T18:52:47.355801Z","shell.execute_reply":"2023-08-03T18:52:47.355827Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if is_inference:\n    model.eval()\n    submission_file = open(\"submission.csv\", \"w\")\n    submission_file.write(\"Id,Predicted\\n\")\n\n    results: list[str] = []\n    \n    for i, data in enumerate(tqdm(test_loader)):\n        res = run_inference(data)\n        results.extend(res)\n        \n        if i % (500 // BATCH) == 0:\n            print(f\"Inference on batch {i}/{len(test_loader)} done\")\n            submission_file.writelines(results)\n            results = []\n\n    submission_file.writelines(results)\n    submission_file.close()","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.357659Z","iopub.status.idle":"2023-08-03T18:52:47.358475Z","shell.execute_reply.started":"2023-08-03T18:52:47.358204Z","shell.execute_reply":"2023-08-03T18:52:47.358227Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"if Path(\"submission.csv\").exists:\n    display(FileLink(\"submission.csv\"))","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.359886Z","iopub.status.idle":"2023-08-03T18:52:47.360714Z","shell.execute_reply.started":"2023-08-03T18:52:47.360428Z","shell.execute_reply":"2023-08-03T18:52:47.360451Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"!rm -r detectron2/","metadata":{"execution":{"iopub.status.busy":"2023-08-03T18:52:47.362246Z","iopub.status.idle":"2023-08-03T18:52:47.363090Z","shell.execute_reply.started":"2023-08-03T18:52:47.362835Z","shell.execute_reply":"2023-08-03T18:52:47.362860Z"},"trusted":true},"execution_count":null,"outputs":[]}]}