3rd Place Solution for Short-video Face Parsing Challenge

TL;DR

This paper presents EANet, an edge-aware network for face parsing in short videos, utilizing edge information and attention loss to improve segmentation accuracy and boundary smoothness, achieving third place in a CVPR challenge.

Contribution

Introduction of EANet with edge attention loss for improved face parsing accuracy and boundary refinement in short-video face segmentation.

Findings

Achieved 86.16% accuracy in face parsing.

Ranked third in the CVPR 2021 challenge.

Effective edge refinement improves segmentation quality.

Abstract

This is a short technical report introducing the solution of Team Rat for Short-video Parsing Face Parsing Track of The 3rd Person in Context (PIC) Workshop and Challenge at CVPR 2021. In this report, we propose an Edge-Aware Network (EANet) that uses edge information to refine the segmentation edge. To further obtain the finer edge results, we introduce edge attention loss that only compute cross entropy on the edges, it can effectively reduce the classification error around edge and get more smooth boundary. Benefiting from the edge information and edge attention loss, the proposed EANet achieves 86.16\% accuracy in the Short-video Face Parsing track of the 3rd Person in Context (PIC) Workshop and Challenge, ranked the third place.