Facial region extraction based on dual-scale detection in 3D face identification

In 3D face recognition tasks, effectively identifying the direction of eye gaze becomes extremely difficult when the interacting user is at a distance. Estimating facial orientation effectively is crucial for determining the regions of interest for the interacting user on the interface, which is another key aspect in achieving context-aware intelligent humancomputer interaction. We have designed a facial region extraction algorithm based on dual-scale detection to accurately locate the facial region of a person. Compared to baselines, our proposed method performs the best and can accurately estimate the facial orientation even in complex scenarios.