Adaptive Weighted Face Alignment by Multi-Scale Feature and Offset Prediction

Table 1

Comparison of results with the state-of-the-art method on the WFLW dataset, error (NME) normalized by the inter-ocular distance.

Method	Test	Pose	Expression	Illumination	Makeup	Occlusion	Blur
LBF [5]	10.29	24.10	11.45	9.32	9.38	13.03	11.28
ESR [30]	11.13	25.88	11.47	10.49	11.05	13.75	12.20
CFSS [4]	9.07	21.36	10.09	8.30	8.74	11.76	9.96
DVLN [31]	6.08	11.54	6.78	5.73	5.98	7.33	6.88
LAB [9]	5.27	10.24	5.51	5.23	5.15	6.79	6.32
Wing [32]	5.11	8.75	5.36	4.93	5.41	6.37	5.81
FCDN [33]	4.86	7.81	5.13	4.77	4.77	5.78	5.34
PIPNet [6]	4.79	8.76	4.86	4.56	4.60	6.04	5.49
MAttHG [21]	4.68	8.18	4.83	4.48	4.71	5.26	5.72
SD-HRNet [34]	4.93	8.63	5.31	4.81	4.76	5.73	5.56
AWMOP	4.68	8.18	4.81	4.54	4.41	5.79	5.38

Bold values represent the best results.