Constrained Joint Cascade Regression Framework for Simultaneous Facial Action Unit Recognition and Facial Landmark Detection

摘要

Cascade regression framework has been shown to be effective for faciallandmark detection. It starts from an initial face shape and gradually predictsthe face shape update from the local appearance features to generate the faciallandmark locations in the next iteration until convergence. In this paper, weimprove upon the cascade regression framework and propose the Constrained JointCascade Regression Framework (CJCRF) for simultaneous facial action unitrecognition and facial landmark detection, which are two related face analysistasks, but are seldomly exploited together. In particular, we first learn therelationships among facial action units and face shapes as a constraint. Then,in the proposed constrained joint cascade regression framework, with the helpfrom the constraint, we iteratively update the facial landmark locations andthe action unit activation probabilities until convergence. Experimentalresults demonstrate that the intertwined relationships of facial action unitsand face shapes boost the performances of both facial action unit recognitionand facial landmark detection. The experimental results also demonstrate theeffectiveness of the proposed method comparing to the state-of-the-art works.