torchvision 0.17.2

image and video datasets and models for torch deep learning

Stars: 15332, Watchers: 15332, Forks: 6836, Open Issues: 985

The pytorch/vision repo was created 7 years ago and the last code push was 4 hours ago.
The project is extremely popular with a mindblowing 15332 github stars!

How to Install torchvision

You can install torchvision using pip

pip install torchvision

or add it to a project with poetry

poetry add torchvision

Package Details

Author

PyTorch Core Team

License

BSD

Homepage

https://github.com/pytorch/vision

PyPi:

https://pypi.org/project/torchvision/

GitHub Repo:

https://github.com/pytorch/vision

Related Packages

Errors

A list of common torchvision errors.

Code Examples

Here are some torchvision code examples and snippets.

GitHub Issues

The torchvision package has 985 open issues on GitHub

• fix bug when the target is empty
• FCOS empty box images
• Replace Assert statement with ValueError message.
• changed assert to ValueError in backbone_utils
• Add support for get_metadata() in GPU decoder
• [WIP] Adding more ConvNeXt variants
• video classification experiments using GPU decoder
• video classification reference script with GPU decoder support
• Add kinetics dataset to use new video reading API
• Add support for get_metadata() in GPU decoder
• Vectorize box decoding in FCOS model
• [PoC] add test to check available feature types of prototype datasets
• Change CPU decoder output frames to use ITU709 colour space
• Allow more consistent degrees=None in RandomAffine constructor
• [feature request] Derive all transforms classes from nn.Module?

See more issues on GitHub