[NeurIPS 2023] This is the official repository for our paper: MPVSS: Mask Propagation for Efficient Video Semantic Segmentation by Yuetian Weng, Mingfei Han, Haoyu He, Mingjie Li, Lina Yao, Xiaojun Chang and Bohan Zhuang.
We have presented a simple yet effective mask propagation framework, dubbed MPVSS, for efficient VSS. Specifically, we have employed a strong query-based image segmentor to process key frames and generate accurate binary masks and class predictions. Then we have proposed to estimate specific flow maps for each segment-level mask prediction of the key frame. Finally, the mask predictions from key frames were subsequently warped to other non-key frames via the proposed query-based flow maps.
See installation instructions for mask2former.
- Download vspw dataset from https://www.vspwdataset.com/
- Create link to the dataset
ln -s /path/to/your/dataset datasets/vspwsh run.sh| Backbone | mIoU | WIoU | VC_8 | VC_16 | GFLOPs | #Params | FPS |
|---|---|---|---|---|---|---|---|
| R50 | 37.5 | 59.0 | 84.1 | 77.2 | 38.9 | 84.1 | 33.93 |
| R101 | 38.8 | 59.0 | 84.8 | 79.6 | 45.1 | 103.1 | 32.38 |
| Swin-T | 39.9 | 62.0 | 85.9 | 80.4 | 39.7 | 114.0 | 32.86 |
| Swin-S | 40.4 | 62.0 | 86.0 | 80.7 | 47.3 | 108.0 | 30.61 |
| Swin-B | 52.6 | 68.4 | 89.5 | 85.9 | 61.5 | 147.0 | 27.38 |
| Swin-L | 53.9 | 69.1 | 89.6 | 85.8 | 97.3 | 255.4 | 23.22 |
| Backbone | mIoU | GFLOPs | #Params (M) | FPS |
|---|---|---|---|---|
| R50 | 78.4 | 173.2 | 84.1 | 13.43 |
| R101 | 78.2 | 204.3 | 103.1 | 12.55 |
| Swin-T | 80.7 | 175.9 | 114.0 | 12.33 |
| Swin-S | 81.3 | 213.2 | 108.0 | 10.98 |
| Swin-B | 81.7 | 278.6 | 147.0 | 9.54 |
| Swin-L | 81.6 | 449.5 | 255.4 | 7.24 |
If you find this repository or our paper useful, please consider cite:
@inproceedings{weng2023mask,
title={Mask Propagation for Efficient Video Semantic Segmentation},
author={Weng, Yuetian and Han, Mingfei and He, Haoyu and Li, Mingjie and Yao, Lina and Chang, Xiaojun and Zhuang, Bohan},
booktitle={NeurIPS},
year={2023}
}
The code is largely based on Mask2Former. We thank the authors for their open-sourced code.