Skip to main content
. 2016 Sep 24;122(1):17–33. doi: 10.1007/s11263-016-0950-1

Fig. 6.

Fig. 6

Fitting performance of pixel-based AAMs and GN-DPMs on LFPW, Helen and AFW: Average pt-pt Euclidean error (normalized by the face size) versus fraction of images. 68 points were used