Shuffle Transformer with Feature Alignment for Video Face Parsing
This is an incremental improvement for video face parsing in computer vision, specifically for challenge performance.
The paper tackles video face parsing by introducing a Shuffle Transformer backbone and Feature Alignment Aggregation module to improve segmentation accuracy, achieving 86.9519% score and first place in the CVPR 2021 PIC Challenge.
This is a short technical report introducing the solution of the Team TCParser for Short-video Face Parsing Track of The 3rd Person in Context (PIC) Workshop and Challenge at CVPR 2021. In this paper, we introduce a strong backbone which is cross-window based Shuffle Transformer for presenting accurate face parsing representation. To further obtain the finer segmentation results, especially on the edges, we introduce a Feature Alignment Aggregation (FAA) module. It can effectively relieve the feature misalignment issue caused by multi-resolution feature aggregation. Benefiting from the stronger backbone and better feature aggregation, the proposed method achieves 86.9519% score in the Short-video Face Parsing track of the 3rd Person in Context (PIC) Workshop and Challenge, ranked the first place.