Pixel-in-Pixel Net: Towards Efficient Facial Landmark Detection in the Wild

Benchmark Model Rank Results
face-alignment-on-300wPIPNet (ResNet-101)#11NME_inter-ocular (%, Full): 3.19NME_inter-ocular (%, Common): 2.78
face-alignment-on-aflw-19PIPNet (ResNet-101)#4NME_diag (%, Full): 1.42
face-alignment-on-cofwPIPNet (ResNet-101)#1NME (inter-ocular): 3.08%
face-alignment-on-wflwPIPNet (ResNet-101)#7NME (inter-ocular): 4.31