Abstract
Short videos have many applications on fashion trends, hot spots, street interviews, public education, and creative advertising. We propose an Edge-Aware Network(EANet) that uses edge information to refine the segmentation edge. And experiments show our proposed EANet boots up the facial parsing results. We also use post-process like grab cut to refine and merge the parsing results.
Abstract (translated)
URL
https://arxiv.org/abs/2106.07409