# What changed in torchvision

> 10 changes in torchvision 0.28 to 0.29 from its official release notes, upgrade guides and advisories: 1 deprecations, 1 security, 1 new APIs and options. Each one cited. Sources checked 2026-10-18. HTML: https://unstale.sh/pypi/torchvision/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## torchvision 0.29

### Deprecations

- The image decoders and encoders in torchvision.io are deprecated and will be removed in a future release; they are available in TorchCodec. (0.29.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.0)

### Behavior and default changes

- The PyTorch requirement is relaxed from an exact 2.14.0 pin to torch>=2.14 because TorchVision is ABI stable with respect to PyTorch 2.14. (0.29.1) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.1)
- TorchVision 0.29 is ABI stable with torch 2.14 and compatible with future torch versions, so upgrading torch will not require a new TorchVision version. (0.29.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.0)

### Fixes and other changes

- Fixed deformable convolution kernels always running on the default stream. (0.29.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.0)
- Fixed the JPEG transform for non-contiguous batches. (0.29.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.0)
- Fixed gradient overaccumulation in ROI ops. (0.29.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.29.0)

## torchvision 0.28

### Security

- Fixes a malformed-GIF decoder bug that could write outside the allocated tensor's memory. (0.28.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.28.0)

### New APIs and options

- Resize transforms now accept strings for the interpolation parameter. (0.28.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.28.0)

### Behavior and default changes

- wrap() preserves metadata for custom TVTensor subclasses. (0.28.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.28.0)

### Fixes and other changes

- F.resize now honors NEAREST_EXACT for tv_tensors.Mask inputs instead of silently using NEAREST. (0.28.0) — [release notes](https://github.com/pytorch/vision/releases/tag/v0.28.0)