SADRNet: Self-Aligned Dual Face Regression Networks for Robust 3D Dense Face Alignment and Reconstruction

6 Jun 2021  ·  Zeyu Ruan, Changqing Zou, Longhai Wu, Gangshan Wu, LiMin Wang ·

Three-dimensional face dense alignment and reconstruction in the wild is a challenging problem as partial facial information is commonly missing in occluded and large pose face images. Large head pose variations also increase the solution space and make the modeling more difficult. Our key idea is to model occlusion and pose to decompose this challenging task into several relatively more manageable subtasks. To this end, we propose an end-to-end framework, termed as Self-aligned Dual face Regression Network (SADRNet), which predicts a pose-dependent face, a pose-independent face. They are combined by an occlusion-aware self-alignment to generate the final 3D face. Extensive experiments on two popular benchmarks, AFLW2000-3D and Florence, demonstrate that the proposed method achieves significant superior performance over existing state-of-the-art methods.

PDF Abstract

Results from the Paper

Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Head Pose Estimation AFLW2000 SADRNet MAE 3.82 # 4
3D Face Reconstruction AFLW2000-3D SADRNet Mean NME 3.25% # 1
Face Alignment AFLW2000-3D SADRNet Mean NME 3.46% # 5
3D Face Reconstruction REALY SADRNet @nose 1.791 (±0.542) # 9
@mouth 1.591 (±0.488) # 8
@forehead 2.413 (±0.537) # 12
@cheek 1.856 (±0.701) # 18
all 1.913 # 9
3D Face Reconstruction REALY (side-view) SADRNet @nose 1.771 (±0.521) # 7
all 1.958 # 8
@mouth 1.560 (±0.462) # 6
@forehead 2.490 (±0.566) # 12
@cheek 2.010 (±0.715) # 16


No methods listed for this paper. Add relevant methods here