175745.fig.0012
Figure 12: Recall versus precision for face detection, face tracking, and speaker match (Dataset 1). Both face tracking and speaker match show good performance as there are only two participants within a sector of 100°.