Face Alignment Across Large Poses: A 3D Solution

Benchmark Model Rank Results
3d-face-reconstruction-on-aflw2000-3d3DDFA–Mean NME: 5.3695%
3d-face-reconstruction-on-florence3DDFA–Mean NME: 6.3833%
face-alignment-on-aflw20003DDFA–Error rate: 5.42
face-alignment-on-aflw2000-3d3DDFA + SDM–Balanced NME (2D Sparse Alignment): 4.94%
facial-landmark-detection-on-300w3DDFA–NME: 7.01
facial-landmark-detection-on-300wCFSS–NME: 5.76
head-pose-estimation-on-aflw20003DDFA–MAE: 7.393
head-pose-estimation-on-biwi3DDFA–MAE (trained with other data): 19.068