Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression

16 Apr 2019Xinyao WangLiefeng BoLi Fuxin

Heatmap regression has became one of the mainstream approaches to localize facial landmarks. As Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) are becoming popular in solving computer vision tasks, extensive research has been done on these architectures... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Face Alignment 300W AWing (Inter-ocular Norm) Mean Error Rate private 3.56 # 2
Face Alignment 300W AWing (Inter-ocular Norm) AUC0.08 private 64.40 # 3
Face Alignment 300W AWing (Inter-ocular Norm) Failure private 0.33% # 2
Face Alignment 300W AWing (Inter-ocular Norm) Fullset (public) 3.07 # 2
Face Alignment 300W AWing (Inter-pupil Norm) Mean Error Rate private 3.56 # 2
Face Alignment 300W AWing (Inter-pupil Norm) AUC0.08 private 55.76 # 5
Face Alignment 300W AWing (Inter-pupil Norm) Failure private 0.83% # 3
Face Alignment 300W AWing (Inter-pupil Norm) Fullset (public) 4.31 # 8
Face Alignment COFW AWing Mean Error Rate 4.94 # 1