Datapoints

Datapoints are tensor subclasses which the v2 v2 transforms use under the hood to dispatch their inputs to the appropriate lower-level kernels. Most users do not need to manipulate datapoints directly and can simply rely on dataset wrapping - see e.g. Transforms v2: End-to-end object detection example.

Image(data, *[, dtype, device, requires_grad])	[BETA] torch.Tensor subclass for images.
Video(data, *[, dtype, device, requires_grad])	[BETA] torch.Tensor subclass for videos.
BoundingBoxFormat(value)	[BETA] Coordinate format of a bounding box.
BoundingBox(data, *, format, spatial_size[, ...])	[BETA] torch.Tensor subclass for bounding boxes.
Mask(data, *[, dtype, device, requires_grad])	[BETA] torch.Tensor subclass for segmentation and detection masks.