Facial Shape Tracking via Spatio-Temporal Cascade Shape Regression

In this paper, we develop a spatio-temporal cascade shape regression (STCSR) model for robust facial shape tracking. It is different from previous works in three aspects. Firstly, a multi-view cascade shape regression (MCSR) model is employed to decrease the shape variance in shape regression model construction, which is able to make the learned regression model more robust to shape variances. Secondly, a time series regression (TSR) model is explored to enhance the temporal consecutiveness between adjacent frames. Finally, a novel re-initialization mechanism is adopted to effectively and accurately locate the face when it is misaligned or lost. Extensive experiments on the 300 Videos in the Wild (300-VW) demonstrate the superior performance of our algorithm.