Two-Stage Convolutional Part Heatmap Regression for the 1st 3D Face Alignment in the Wild (3DFAW) Challenge

被引:49
作者
Bulat, Adrian [1 ]
Tzimiropoulos, Georgios [1 ]
机构
[1] Univ Nottingham, Comp Vision Lab, Nottingham, England
来源
COMPUTER VISION - ECCV 2016 WORKSHOPS, PT II | 2016年 / 9914卷
基金
英国工程与自然科学研究理事会;
关键词
3D face alignment; Convolutional Neural Networks; Convolutional part heatmap regression;
D O I
10.1007/978-3-319-48881-3_43
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
This paper describes our submission to the 1st 3D Face Alignment in the Wild (3DFAW) Challenge. Our method builds upon the idea of convolutional part heatmap regression (Bulat and Tzimiropoulos, 2016), extending it for 3D face alignment. Our method decomposes the problem into two parts: (a) X, Y (2D) estimation and (b) Z (depth) estimation. At the first stage, our method estimates the X, Y coordinates of the facial landmarks by producing a set of 2D heatmaps, one for each landmark, using convolutional part heatmap regression. Then, these heatmaps, alongside the input RGB image, are used as input to a very deep subnetwork trained via residual learning for regressing the Z coordinate. Our method ranked 1st in the 3DFAW Challenge, surpassing the second best result by more than 22%.
引用
收藏
页码:616 / 624
页数:9
相关论文
共 21 条
  • [1] [Anonymous], CVPR W
  • [2] [Anonymous], IMAGE VISION COMPUTI
  • [3] Human Pose Estimation via Convolutional Part Heatmap Regression
    Bulat, Adrian
    Tzimiropoulos, Georgios
    [J]. COMPUTER VISION - ECCV 2016, PT VII, 2016, 9911 : 717 - 732
  • [4] Bulat Adrian, 2016, BMVC
  • [5] Cao X., 2012, CVPR
  • [6] Collobert R, 2011, BIGLEARN NIPS WORKSH, P1
  • [7] Deng J, 2009, PROC CVPR IEEE, P248, DOI 10.1109/CVPRW.2009.5206848
  • [8] Multi-PIE
    Gross, Ralph
    Matthews, Iain
    Cohn, Jeffrey
    Kanade, Takeo
    Baker, Simon
    [J]. IMAGE AND VISION COMPUTING, 2010, 28 (05) : 807 - 813
  • [9] He K., 2016, P IEEE C COMPUTER VI, P770, DOI DOI 10.1109/CVPR.2016.90
  • [10] He K., 2016, PROC CVPR IEEE, P630, DOI [10.1007/978-3-319-46493-0_38, DOI 10.1007/978-3-319-46493-0_38, DOI 10.1109/CVPR.2016.90]