1 Release published by 1 person

• v0.22.0 Torchvision 0.22 release

  published Apr 23, 2025

9 Pull requests merged by 5 people

• Fix Linux aarch64 nightly builds. Set with-cuda: enabled

  #9062 merged May 9, 2025

• [release-only] Bump version to 0.22.1

  #9061 merged May 9, 2025

• Update .git-blame-ignore-revs

  #9035 merged Apr 23, 2025

• Upgrade type hint and others to Python 3.9

  #8814 merged Apr 23, 2025

• Revert "[Manylinux_2_28] Modify relocate script for linux builds: fix…

  #9032 merged Apr 22, 2025

• [Build] Add homebrew as library search location

  #9027 merged Apr 22, 2025

• [Manylinux_2_28] Modify relocate script for linux builds: fix wheel_name

  #9024 merged Apr 22, 2025

• Add label background

  #9018 merged Apr 16, 2025

• Fix rotated box format conversion from XYXYXYXY to XYWHR

  #9019 merged Apr 16, 2025

6 Pull requests opened by 6 people

• [MPS] deformable conv2d kernel

  #9017 opened Apr 14, 2025

• Deform conv2d mps support

  #9026 opened Apr 20, 2025

• Raise error when download=True for LFW dataset #8888

  #9040 opened Apr 26, 2025

• Fix #9051: Add integer dtype check to RandomPosterize

  #9052 opened May 2, 2025

• Batched Box Ops

  #9058 opened May 4, 2025

• Removing Conda references.

  #9063 opened May 12, 2025

14 Issues closed by 6 people

• torchvision 0.22.0+cu128 / torchvision.utils.save_image causes RuntimeError

  #9047 closed Apr 30, 2025

• CocoDetection super().attr Inheritance Results in AttributeError

  #9041 closed Apr 28, 2025

• Batching tv_tensors.Image tensors does not preserve tv_tensors.Image type

  #9039 closed Apr 28, 2025

• Setting `NEAREST_EXACT` to `interpolation` argument of `RandomRotation()` works even though error messages don't say

  #9038 closed Apr 25, 2025

• `io.video.write_video()` has `fps: float` but underlying `add_stream()` from `av` expects integers or fractions

  #9037 closed Apr 23, 2025

• Setting `InterpolationMode.TRILINEAR` and `InterpolationMode.AREA` to `interpolation` argument of `Resize()` for a tensor still gets errors against an error messasge. *The error message should be incorrect.

  #9034 closed Apr 23, 2025

• `ToDtype()` also scales tensors

  #9033 closed Apr 23, 2025

• Use clang to compile torch 2.3.1 and torchvision 0.18 ,got error

  #9031 closed Apr 22, 2025

• `ToDtype()` also converts the dtypes of the tensors of `int`, `complex` and `bool` to `float`

  #9030 closed Apr 22, 2025

• Setting the `np.array([[1000.0, 2000.0]]` which isn't (H x W x C) in the range `[0, 255]` to `ToTensor()` is scaled in the range `[0.0, 1.0]` against the doc

  #9028 closed Apr 22, 2025

• With `p=0`, setting a list of no transforms to `transforms` argument of `RandomApply()` doesn't get error

  #9023 closed Apr 22, 2025

• write_video documentation's sample isn't working

  #9021 closed Apr 22, 2025

• Apple Clang 16 build issue, requires ld_classic

  #8653 closed Apr 17, 2025

• ops.nms get wrong results when set iou_threshold=0

  #8949 closed Apr 14, 2025

21 Issues opened by 9 people

• encode_jpeg generates corrupted JPEG data when using CUDA

  #9060 opened May 7, 2025

• `download=True` argument for `VOCDetection` will always download and rewrite the dataset again

  #9059 opened May 5, 2025

• Setting a `float` type tensor to `AugMix()` works against the error messages

  #9057 opened May 2, 2025

• Setting a `float` type tensor to `AutoAugment()` works against the error messages

  #9056 opened May 2, 2025

• Setting a `float` type tensor to `TrivialAugmentWide()` works against the error messages

  #9055 opened May 2, 2025

• Setting a `float` type tensor to `RandomSolarize()` works against the error message

  #9054 opened May 2, 2025

• Setting a `float` type tensor to `RandomInvert()` works against the error messages

  #9053 opened May 2, 2025

• Setting a `float` type tensor to `RandomPosterize()` works against the error messages

  #9051 opened May 2, 2025

• Add Unlabeled Image Dataset for Unsupervised Training

  #9050 opened May 2, 2025

• Setting a `float` type tensor to `RandomEqualize()` works against the error messages

  #9049 opened May 2, 2025

• Setting a 0D or 1D tensor to `RandomPerspective()` sometimes works against the error message

  #9048 opened May 2, 2025

• ssd lite models cannot use cocoeval such as map and ap but faster rccn models can run well.

  #9046 opened Apr 30, 2025

• Setting the tensors and ndarrays of the 2, 3 and 4 channels of `bool` to `ToPILImage()` with `None` for `mode` argument gets errors

  #9045 opened Apr 30, 2025

• Setting the tensors and ndarrays of the 2, 3 and 4 channels of `complex64` and `complex128` to `ToPILImage()` with `None` for `mode` argument gets errors

  #9044 opened Apr 30, 2025

• Setting the tensors and ndarrays of the 2, 3 and 4 channels of `int64` and `int32` to `ToPILImage()` with `None` for `mode` argument gets errors

  #9043 opened Apr 30, 2025

• Make the C++ backend of the torchvision wheel usable for C++ development

  #9042 opened Apr 29, 2025

• Update on TorchVision’s future focus

  #9036 opened Apr 23, 2025

• Rotation sampling grid is undesireably low-res

  #9029 opened Apr 22, 2025

• Setting a 1D tensor to `ToImage()` gets an empty error message

  #9025 opened Apr 19, 2025

• Setting a list of no transforms to `transforms` argument of `RandomApply()` gets the indirect error message

  #9022 opened Apr 18, 2025

• AttributeError: module 'torch.library' has no attribute 'register_fake'

  #9020 opened Apr 16, 2025

16 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Added the KeyPoints TVTensor

  #8817 commented on May 5, 2025 • 51 new comments

• Add `box_area_center` and `box_iou_center` functions for cxcywh format with tests

  #8992 commented on May 1, 2025 • 1 new comment

• `convert_image_dtype` overflows with low precision floating point dtypes

  #6799 commented on Apr 14, 2025 • 0 new comments

• position, colour, and background colour of text labels in draw_bounding_boxes

  #8317 commented on Apr 17, 2025 • 0 new comments

• deform_conv2d for mps

  #7490 commented on Apr 20, 2025 • 0 new comments

• optests.generate_opcheck_tests

  #8731 commented on Apr 21, 2025 • 0 new comments

• LFW dataset support

  #8888 commented on Apr 22, 2025 • 0 new comments

• Speed up JPEG decoding by allowing resize during decode

  #8986 commented on Apr 29, 2025 • 0 new comments

• box_area and box_iou functions for cxcywh format

  #8961 commented on Apr 29, 2025 • 0 new comments

• [feature request] Image Histogram Transformation

  #598 commented on May 5, 2025 • 0 new comments

• Setting `0` and `1` to `p` argument of `RandomAutocontrast()` gets the same results

  #8915 commented on May 7, 2025 • 0 new comments

• [NOMRG] Current CI status

  #8601 commented on Apr 15, 2025 • 0 new comments

• Improve type hints in tv_tensor.wrap

  #8830 commented on Apr 30, 2025 • 0 new comments

• Arm64 workflow enablement

  #8947 commented on May 8, 2025 • 0 new comments

• Draft - Test Workflow arm64 logic

  #8948 commented on May 2, 2025 • 0 new comments

• Fix the PhotoTour download URLs

  #9002 commented on Apr 14, 2025 • 0 new comments