Look at Boundary: A Boundary-Aware Face Alignment Algorithm

Abstract

We present a novel boundary-aware face alignment algorithm by utilising boundary lines as the geometric structure of a human face to help facial landmark localisation. Unlike the conventional heatmap based method and regression based method, our approach derives face landmarks from boundary lines which remove the ambiguities in the landmark definition. Three questions are explored and answered by this work: 1. Why using boundary? 2. How to use boundary? 3. What is the relationship between boundary estimation and landmarks localisation? Our boundaryaware face alignment algorithm achieves 3.49% mean error on 300-W Fullset, which outperforms state-of-the-art methods by a large margin. Our method can also easily integrate information from other datasets. By utilising boundary information of 300-W dataset, our method achieves 3.92% mean error with 0.39% failure rate on COFW dataset, and 1.25% mean error on AFLW-Full dataset. Moreover, we propose a new dataset WFLW to unify training and testing across different factors, including poses, expressions, illuminations, makeups, occlusions, and blurriness. Dataset and model will be publicly available at https://wywu. github.io/projects/LAB/LAB.html

DeCaFA: Deep Convolutional Cascade for Face Alignment in the Wild

“…We also report the error on multiple test subsets containing variations in head pose, facial expressions, illumination, make-up as well as partial occlusions and occasional blur. DeCaFA performs better than LAB [19] and Wing [6] by a significant margin on every subset. Also, note that DeCaFA trained solely on WFLW already as a ME of 5.01 on the whole test set, which is still better that these two methods.…”

“…This aggregated style space thus serve as an intermediate representation that is more convenient for training. In [19] the authors propose to use edge map estimation as an intermediate representation to drive the landmark prediction task, as well as to provide a unified representation when images are annotated in terms of different markups, that correspond to different alignment tasks. Finally, DSRN [13] relies on Fourier Embedding and low-rank learning to produce such representation.…”

FAB: A Robust Facial Landmark Detection Framework for Motion-Blurred Videos

“…It currently contains 20 videos with obvious real-world motion blur picked from YouTube, which include dancing, boxing, jumping, etc. There are 35, 540 frames, which are all annotated with 98 landmarks following the protocol of WFLW [38]. Moreover, RWMB dataset will be further enlarged to hundreds of videos, including millions of frames.…”

“…We provided two versions of annotation of RWMB and 300VW, i.e. 98 landmarks following [38] and 68 landmarks following [34]. In this way, we…”

“…This experiment is conducted to demonstrate that our algorithm is capable to handle artificial motion blur, and accurately detect facial landmarks. We re-implemented several typical state-of-the-art algorithms for comparison, including RFLD [25], SAN [8] and LAB [38]. For a fair comparison, we made great efforts to reproduce the comparable results reported in their papers.…”