Multi-task head pose estimation in-the-wild

Benchmark Model Rank Results
face-alignment-on-aflw2000MNN+ORB (Reannotated)#1Error rate: 2.58
face-alignment-on-aflw2000-3dMNN+OR (reannotated)#1Balanced NME (2D Sparse Alignment): 2.58%
face-alignment-on-cofwMNN+OR (Inter-pupils Norm)#15Recall at 80% precision (Landmarks Visibility): 72.12
face-alignment-on-cofwMNN (Inter-pupil Norm)#20NME (inter-pupil): 5.65%
head-pose-estimation-on-aflwMNN#1MAE: 3.22
head-pose-estimation-on-aflw2000MNN#9MAE: 3.83
head-pose-estimation-on-biwiMNN#9MAE (trained with other data): 3.66
pose-estimation-on-300w-fullMNN#2MAE mean (º): 1.56