Common view image

Hello, I have the honor to read your article. I would like to ask how to get the common view image after the mask as shown in Figure 4 of the paper according to the mask matrix.

Hi, thanks for your interest in our work!
Do you mean how do I obtain mask? To obtain the mask we first estimate the a dense correspondence map from image A to image B. We then feed the predicted target locations in image B as the query points, to get another dense correspondence map from B to A. It allows us to compute the cycle consistency, and we simply mask out areas with high cycle consistency error.
If our network is perfect, the Target=f(Query) and f(f(Query))=Query, and it means 0 cycle consistency error.
You can check these two lines to see how we obtain the masks:

COTR/COTR/inference/sparse_engine.py

Lines 142 to 143 in 21c31ed

    
           mask_a = con_a < THRESHOLD_SPARSE 
        
           mask_b = con_b < THRESHOLD_SPARSE

	mask_a = con_a < THRESHOLD_SPARSE
	mask_b = con_b < THRESHOLD_SPARSE