ZeroCo is a zero-shot correspondence model that demonstrates the effectiveness of cross-attention maps, learned through cross-view completion training, in capturing correspondences. In this work, we ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results