Abstract
Existing methods of 3D dense face alignment mainly concentrate on accuracy,thus limiting the scope of their practical applications. In this paper, wepropose a novel regression framework which makes a balance among speed,accuracy and stability. Firstly, on the basis of a lightweight backbone, wepropose a meta-joint optimization strategy to dynamically regress a small setof 3DMM parameters, which greatly enhances speed and accuracy simultaneously.To further improve the stability on videos, we present a virtual synthesismethod to transform one still image to a short-video which incorporatesin-plane and out-of-plane face moving. On the premise of high accuracy andstability, our model runs at over 50fps on a single CPU core and outperformsother state-of-the-art heavy models simultaneously. Experiments on severalchallenging datasets validate the efficiency of our method. Pre-trained modelsand code are available at https://github.com/cleardusk/3DDFA_V2.