Figure 4.

Example frames from videos (V${}_0$–V${}_3$), used to evaluate HPE-CNNs. Each image is representative of the special aspect of the respective video. (V${}_0$) is a standard court situation. (V${}_1$) shows reflections on the side walls (i) and (V${}_2$) additionally shows the mirrored audience (ii). In (V${}_3$), a supporting beam (iii) occludes the image. Please note: due to privacy reasons, players were removed and names, scores and sponsors were blurred.