SAMatcher

SAMatcher is a robust feature matching method that models cross-view co-visibility with Segment Anything (SAM), improving correspondence under viewpoint/illumination/texture changes.

• GitHub: https://github.com/TwSphinx54/SAMatcher
• Project Page: https://xupan.top/Projects/samatcher

Quick Start

huggingface-cli download SSSSSphinx/SAMatcher --local-dir ./SAMatcher

Then follow the official training/evaluation scripts in the GitHub repository.

Notes

• Recommended Python: 3.12 (if using lockfile setup)
• Main dependencies include PyTorch + CUDA 12.1 and flash-attn