From 9bb7e1831b0cb11717a7f6dc67a9ba5fbe46d163 Mon Sep 17 00:00:00 2001 From: Philip Meier Date: Mon, 27 Feb 2023 13:29:19 +0100 Subject: [PATCH 1/3] add docstring for dataset wrapper (#7333) Co-authored-by: Nicolas Hug --- docs/source/datasets.rst | 9 +++ torchvision/datapoints/_dataset_wrapper.py | 65 +++++++++++++++++++++- 2 files changed, 73 insertions(+), 1 deletion(-) diff --git a/docs/source/datasets.rst b/docs/source/datasets.rst index 68c72e7af8c..35e5eaf2a9f 100644 --- a/docs/source/datasets.rst +++ b/docs/source/datasets.rst @@ -169,3 +169,12 @@ Base classes for custom datasets DatasetFolder ImageFolder VisionDataset + +Transforms v2 +------------- + +.. autosummary:: + :toctree: generated/ + :template: function.rst + + wrap_dataset_for_transforms_v2 diff --git a/torchvision/datapoints/_dataset_wrapper.py b/torchvision/datapoints/_dataset_wrapper.py index e358c83d9d1..87ce3ba93a1 100644 --- a/torchvision/datapoints/_dataset_wrapper.py +++ b/torchvision/datapoints/_dataset_wrapper.py @@ -14,8 +14,71 @@ __all__ = ["wrap_dataset_for_transforms_v2"] -# TODO: naming! def wrap_dataset_for_transforms_v2(dataset): + """[BETA] Wrap a ``torchvision.dataset`` for usage with :mod:`torchvision.transforms.v2`. + + .. v2betastatus:: wrap_dataset_for_transforms_v2 function + + Example: + >>> dataset = torchvision.datasets.CocoDetection(...) + >>> dataset = wrap_dataset_for_transforms_v2(dataset) + + .. note:: + + For now, only the most popular datasets are supported. Furthermore, the wrapper only supports dataset + configurations that are fully supported by ``torchvision.transforms.v2``. If you encounter an error prompting you + to raise an issue to ``torchvision`` for a dataset or configuration that you need, please do so. + + The dataset samples are wrapped according to the description below. + + Special cases: + + * :class:`~torchvision.datasets.CocoDetection`: Instead of returning the target as list of dicts, the wrapper + returns a dict of lists. In addition, the key-value-pairs ``"boxes"`` (in ``XYXY`` coordinate format), + ``"masks"`` and ``"labels"`` are added and wrap the data in the corresponding ``torchvision.datapoints``. + The original keys are preserved. + * :class:`~torchvision.datasets.VOCDetection`: The key-value-pairs ``"boxes"`` and ``"labels"`` are added to + the target and wrap the data in the corresponding ``torchvision.datapoints``. The original keys are + preserved. + * :class:`~torchvision.datasets.CelebA`: The target for ``target_type="bbox"`` is converted to the ``XYXY`` + coordinate format and wrapped into a :class:`~torchvision.datapoints.BoundingBox` datapoint. + * :class:`~torchvision.datasets.Kitti`: Instead returning the target as list of dictsthe wrapper returns a dict + of lists. In addition, the key-value-pairs ``"boxes"`` and ``"labels"`` are added and wrap the data + in the corresponding ``torchvision.datapoints``. The original keys are preserved. + * :class:`~torchvision.datasets.OxfordIIITPet`: The target for ``target_type="segmentation"`` is wrapped into a + :class:`~torchvision.datapoints.Mask` datapoint. + * :class:`~torchvision.datasets.Cityscapes`: The target for ``target_type="semantic"`` is wrapped into a + :class:`~torchvision.datapoints.Mask` datapoint. The target for ``target_type="instance"`` is *replaced* by + a dictionary with the key-value-pairs ``"masks"`` (as :class:`~torchvision.datapoints.Mask` datapoint) and + ``"labels"``. + * :class:`~torchvision.datasets.WIDERFace`: The value for key ``"bbox"`` in the target is converted to ``XYXY`` + coordinate format and wrapped into a :class:`~torchvision.datapoints.BoundingBox` datapoint. + + Image classification datasets + + This wrapper is a no-op for image classification datasets, since they were already fully supported by + :mod:`torchvision.transforms` and thus no change is needed for :mod:`torchvision.transforms.v2`. + + Segmentation datasets + + Segmentation datasets, e.g. :class:`~torchvision.datasets.VOCSegmentation` return a two-tuple of + :class:`PIL.Image.Image`'s. This wrapper leaves the image as is (first item), while wrapping the + segmentation mask into a :class:`~torchvision.datapoints.Mask` (second item). + + Video classification datasets + + Video classification datasets, e.g. :class:`~torchvision.datasets.Kinetics` return a three-tuple containing a + :class:`torch.Tensor` for the video and audio and a :class:`int` as label. This wrapper wraps the video into a + :class:`~torchvision.datapoints.Video` while leaving the other items as is. + + .. note:: + + Only datasets constructed with ``output_format="TCHW"`` are supported, since the alternative + ``output_format="THWC"`` is not supported by :mod:`torchvision.transforms.v2`. + + Args: + dataset: the dataset instance to wrap for compatibility with transforms v2. + """ return VisionDatasetDatapointWrapper(dataset) From 037c00620c95fb2eac14aa3d54b18726d8a50f54 Mon Sep 17 00:00:00 2001 From: Nicolas Hug Date: Thu, 2 Mar 2023 17:13:58 +0000 Subject: [PATCH 2/3] empty commit From a270e363f58411af2aea4ff9b83437b58726d6cb Mon Sep 17 00:00:00 2001 From: Camilo De La Torre <64303300+camilodlt@users.noreply.github.com> Date: Tue, 7 Mar 2023 14:53:41 +0100 Subject: [PATCH 3/3] Fix typo in warning message (#7394) --- torchvision/__init__.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/torchvision/__init__.py b/torchvision/__init__.py index eed24091a52..590b32732ac 100644 --- a/torchvision/__init__.py +++ b/torchvision/__init__.py @@ -105,7 +105,7 @@ def _is_tracing(): "this issue: https://github.com/pytorch/vision/issues/6753, and you can also " "check out https://github.com/pytorch/vision/issues/7319 to learn more about " "the APIs that we suspect might involve future changes. " - "You can silence this warning by calling torchvision.disable_beta_transform_warning()." + "You can silence this warning by calling torchvision.disable_beta_transforms_warning()." )