A Parametric Approach Using Active Shape Models for Facial Landmark Detection

Abstract

On later tasks involving face recognition, three-dimensional face reconstruction, and other related tasks, facial landmark detection is essential. In this study, we introduce a parametric active shape model method for face landmark identification. In particular, we suggest using the active shape model (ASM) parameters to encode the landmark locations. Next, we estimate the ASM parameters by utilizing the power of cascade regression. The ASM parameters can be used to decode final landmark positions. Compared to previous similar work, this parametric method of predicting the landmark placements in another domain is more efficient and compact. We test the efficacy of our approach through experiments.