#493 · Primary category: Computer Vision

FBA_Matting

alphamatting computer-vision deep-image-matting deep-learning deep-matting fba-matting matting pytorch

Official repository for the paper F, B, Alpha Matting

Project last updated:06/22/23

GitHub Stars

502

Forks

96

Contributors

2

License

MIT

Why we included this project

Need to pull a clean subject out of a photo, hair strands and all? FBA Matting predicts an alpha matte plus separate foreground and background colors, which is what makes compositing hair, fur, and translucent objects look convincing. The repo includes pretrained weights, a demo script, a notebook, and a Colab, so you can run the model without training it yourself, and there's a TorchScript export path if you want a more portable deployment. Keep in mind that the released models fall under the Adobe dataset license, so they're restricted to noncommercial use, and the training code was never published. That makes this a solid inference baseline and a useful reference for matting research rather than a complete open pipeline you can retrain end to end.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category