1 paper · 1 filter
Qiang Wang
We revisit the problem of training attention-based sparse image matching models for various local features. We first identify one critical design choice that has been previously ov…