Dense Face Alignment

Yaojie Liu, Amin Jourabloo, William Ren, Xiaoming Liu; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 1619-1628


Face alignment is a classic problem in the computer vision field. Previous research mostly focus on sparse alignment of the face image with a limited set of facial landmark points, i.e., facial landmark detection. In this paper, for the first time, we aim to provide more detailed dense 3D alignment for large-pose face images. To achieve this, we train a deep convolutional network to estimate the 3D shape model parameters, which not only aligns the limited number of facial landmarks, but also fits contours and SIFT feature points and utilizes them as a dense supervision. Moreover, we also address the bottleneck of training with multiple datasets, due to different landmark mark-ups such as 5, 34, 68 and even no labeling. Experimental results show that our method not only provides high-quality dense 3D face fitting, but also outperforms the state-of-the-art facial landmark detection methods on the challenging datasets, with one plain network and at real time.

Related Material

[pdf] [arXiv]
author = {Liu, Yaojie and Jourabloo, Amin and Ren, William and Liu, Xiaoming},
title = {Dense Face Alignment},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}