English

TapLab: A Fast Framework for Semantic Video Segmentation Tapping into Compressed-Domain Knowledge

Computer Vision and Pattern Recognition 2020-08-19 v3

Abstract

Real-time semantic video segmentation is a challenging task due to the strict requirements of inference speed. Recent approaches mainly devote great efforts to reducing the model size for high efficiency. In this paper, we rethink this problem from a different viewpoint: using knowledge contained in compressed videos. We propose a simple and effective framework, dubbed TapLab, to tap into resources from the compressed domain. Specifically, we design a fast feature warping module using motion vectors for acceleration. To reduce the noise introduced by motion vectors, we design a residual-guided correction module and a residual-guided frame selection module using residuals. TapLab significantly reduces redundant computations of the state-of-the-art fast semantic image segmentation models, running 3 to 10 times faster with controllable accuracy degradation. The experimental results show that TapLab achieves 70.6% mIoU on the Cityscapes dataset at 99.8 FPS with a single GPU card for the 1024x2048 videos. A high-speed version even reaches the speed of 160+ FPS. Codes will be available soon at https://github.com/Sixkplus/TapLab.

Keywords

Cite

@article{arxiv.2003.13260,
  title  = {TapLab: A Fast Framework for Semantic Video Segmentation Tapping into Compressed-Domain Knowledge},
  author = {Junyi Feng and Songyuan Li and Xi Li and Fei Wu and Qi Tian and Ming-Hsuan Yang and Haibin Ling},
  journal= {arXiv preprint arXiv:2003.13260},
  year   = {2020}
}

Comments

Accepted to TPAMI

R2 v1 2026-06-23T14:31:28.101Z