Skip to main content

CocoDataPyTorch

CocoDataPyTorch​

CocoDataPyTorch(dataset_name: str = 'coco_validation_with_annotations', root: str = 'coco/val2017', annotationsFile: str = 'coco/annotations/instances_val2017.json')

Bases: PyTorchData

ApertureDB ingestable Dataset, which is sourced from CocoDetection (torchvision.datasets)

COCO dataset loads as an iterable with Tuple (X, [y1, y2 .... yn]) where X is the image (PIL.Image) and y's are multiple dicts with properties like: area, bbox, category_id, image_id, id, iscrowd, keypoints, num_keypoints, segmentation