See this excerpt from the paper. I’m struggling to understand the statement “each dustbin has as many matches as there are keypoints in the other set” from the SuperGlue paper.