The description of Fig. 5 in the CVPR paper is incorrect. From the graph Fig.5, TimeSurface should have the highest mAP with lowest GWD behind ERGO-12. But instead MDES is described as being the best. The issue also appears in the latest arxiv version. Am I reading Fig. 5 incorrectly?
