Combining Local and Global Features for 3D Face Tracking

Pengfei Xiong, Guoqing Li, Yuhang Sun; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2529-2536

This paper presents our framework submitted to 1st 3D Face Tracking in-the-wild Competition. Different from 2d landmark tracking, 3d shapes are more fragile under face posture changes. In order to better capture the various shape and spatial relationships associated with the face, we propose a two stage shape regression method by combining the powerful local heatmap regression and global shape regression. Concretely, stacked hourglass network is adopted to generate a set of heatmaps for each 3d shape point by first. While these heatmaps are independent on each other, a hierarchical attention mechanism is applied from global to local heatmaps into the network, in order to model the correlations among neighboring regions. Extensive experiments on four challenging datasets, show that our proposed algorithm outperforms state-of-the-art baselines.