English

RS-SSM: Refining Forgotten Specifics in State Space Model for Video Semantic Segmentation

Computer Vision and Pattern Recognition 2026-03-27 v2

Abstract

Recently, state space models have demonstrated efficient video segmentation through linear-complexity state space compression. However, Video Semantic Segmentation (VSS) requires pixel-level spatiotemporal modeling capabilities to maintain temporal consistency in segmentation of semantic objects. While state space models can preserve common semantic information during state space compression, the fixed-size state space inevitably forgets specific information, which limits the models' capability for pixel-level segmentation. To tackle the above issue, we proposed a Refining Specifics State Space Model approach (RS-SSM) for video semantic segmentation, which performs complementary refining of forgotten spatiotemporal specifics. Specifically, a Channel-wise Amplitude Perceptron (CwAP) is designed to extract and align the distribution characteristics of specific information in the state space. Besides, a Forgetting Gate Information Refiner (FGIR) is proposed to adaptively invert and refine the forgetting gate matrix in the state space model based on the specific information distribution. Consequently, our RS-SSM leverages the inverted forgetting gate to complementarily refine the specific information forgotten during state space compression, thereby enhancing the model's capability for spatiotemporal pixel-level segmentation. Extensive experiments on four VSS benchmarks demonstrate that our RS-SSM achieves state-of-the-art performance while maintaining high computational efficiency. The code is available at https://github.com/zhoujiahuan1991/CVPR2026-RS-SSM.

Keywords

Cite

@article{arxiv.2603.24295,
  title  = {RS-SSM: Refining Forgotten Specifics in State Space Model for Video Semantic Segmentation},
  author = {Kai Zhu and Zhenyu Cui and Zehua Zang and Jiahuan Zhou},
  journal= {arXiv preprint arXiv:2603.24295},
  year   = {2026}
}

Comments

Accepted by CVPR 2026

R2 v1 2026-07-01T11:37:17.209Z