Get foreground in image without user marking borders using 100x smaller model

Get foreground in image without user marking borders using 100x smaller model

Highly Efficient Natural Image Matting
arXiv paper abstract https://arxiv.org/abs/2110.12748
arXiv PDF paper https://arxiv.org/pdf/2110.12748.pdf

… deep learning based … achieved outstanding improvements in natural image matting.

… still two drawbacks … reliance on user-provided trimaps and the heavy model sizes.

… propose a trimap-free natural image matting method with a lightweight model.

… build a two-stages framework: … to capture sufficient semantics and classify the pixels into unknown, foreground and background regions

… can efficiently model the relevance between different pixels and help regress high-quality alpha values.

… achieves comparable performance with ~1% parameters (344k) of large models on popular natural image matting benchmarks.

Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website

LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b

Photo by Kelly Sikkema on Unsplash