Stylized Avatar Animation Based on Expression Recognition Mapped Deep Learning

The generation of animated character expressions not only requires the reproduction of facial feature point information but also the vivid representation of facial expressions. There are differences in parameter rules between facial expressions and animated character models, and directly generating expressions may lead to exaggerated and unnatural expressions. This article proposes a technique for generating animated character expressions based on facial expression recognition limitations. Firstly, recognize and classify facial expressions based on the deep learning network, and then extract facial feature points. The expression generation model is established by using animated character rig parameters and training on a large amount of data, and the parameter mapping restriction rules are obtained based on facial key points distance. In the process of generating animated expressions, the parameter rules are effectively restricted according to different expression categories, and a script is established to achieve the generation of animated character expressions.