Fig. 8

First frame of each video, with the white shading indicating ignore areas where annotations and estimations are not considered for the computation of the performance measures