Face Alignment(Face Landmark)

state-of-art 2017

http://www.csc.kth.se/~vahidk/face_ert.html

Benchmark*

	STASM	CompASM	EXEM	RCPR	SDM	ESR	ERT (Ours)
LFPW	-	-	0.040	0.035	0.035	0.034**	0.038
HELEN	0.111	0.091	-	0.065	0.059	0.059	0.049
IBUG***	-	-	-	-	0.075	0.075	0.064

STASM, CompASM: Interactive Facial Feature Localization.
EXEM: Localizing Parts of Faces Using a Consensus of Exemplars.
RCPR: Robust Face Landmark Estimation Under Occlusion
SDM: Supervised Descent Method and its Applications to Face Alignment.
ESR: Face Alignment by Explicit Shape Regression.

目前，看到最新的论文，这两个方法的效果基本上等同、甚至超越SDM：

Paper 1:

Cascaded Continuous Regression for Real-time Incremental Face Tracking

eccv2016

https://github.com/ESanchezLozano/iCCR

Paper 2:

Face Alignment by Coarse-to-Fine Shape Searching

cvpr2015

https://github.com/zhusz/CVPR15-CFSS