Facial action unit detection using kernel partial least squares

摘要: In this work, we propose a framework for simultaneously detecting the presence of multiple facial action units using kernel partial least square regression (KPLS). This method has advantage being easily extensible to learn more face related labels, while at same time computationally efficient. We compare approach linear and non-linear support vector machines (SVM) evaluate its performance on extended Cohn-Kanade (CK+) dataset GEneva Multimodal Emotion Portrayals (GEMEP-FERA) dataset, as well across databases. It is shown that KPLS achieves around 2% absolute improvement over SVM-based in terms two alternative forced choice (2AFC) score when trained CK+ tested GEMEP-FERA. 6% GEMEP-FERA CK+. also show handling non-additive AU combinations better than approaches detect single AUs only.