otx.algorithms.visual_prompting.adapters.pytorch_lightning.datasets.pipelines.transforms#

Collection of transfrom pipelines for visual prompting task.

Functions

Collate function for dataloader.

Classes

`MultipleInputsCompose`(transforms)	Composes several transforms have multiple inputs together.
`Pad`()	Pad images, gt_masks, bboxes, and points to the same size.

class otx.algorithms.visual_prompting.adapters.pytorch_lightning.datasets.pipelines.transforms.MultipleInputsCompose(transforms)[source]#

Bases: Compose

Composes several transforms have multiple inputs together.

__call__(item: Dict[str, int | Tensor]) → Dict[str, int | Tensor][source]#

Composes several transforms have multiple inputs together.

Parameters:: item (Dict[str, Union[int, Tensor]]) – Input item.
Returns:: Transformed item.
Return type:: Dict[str, Union[int, Tensor]]

class otx.algorithms.visual_prompting.adapters.pytorch_lightning.datasets.pipelines.transforms.Pad[source]#

Pad images, gt_masks, bboxes, and points to the same size.

__call__(item: Dict[str, List[Any] | Tensor]) → Dict[str, int | Tensor][source]#

Pad images, gt_masks, bboxes, and points to the same size.

Parameters:: item (Dict[str, Union[int, Tensor]]) – Input item.
Returns:: Padded item.
Return type:: Dict[str, Union[int, Tensor]]

otx.algorithms.visual_prompting.adapters.pytorch_lightning.datasets.pipelines.transforms.collate_fn(batch: List[Any]) → Dict[source]#

Collate function for dataloader.