Temporal Cluster Assignment for Efficient Real-Time Video Segmentation

TL;DR

This paper introduces Temporal Cluster Assignment (TCA), a novel, lightweight, and training-free method that leverages temporal coherence to improve token clustering, significantly reducing computation while maintaining accuracy in real-time video segmentation.

Contribution

TCA enhances token clustering by exploiting temporal redundancy, enabling more efficient and accurate real-time video segmentation without additional training.

Findings

TCA improves accuracy-speed trade-off across multiple datasets.

TCA generalizes well to both natural and domain-specific videos.

TCA reduces computational cost significantly while maintaining fine-grained details.

Abstract

Vision Transformers have substantially advanced the capabilities of segmentation models across both image and video domains. Among them, the Swin Transformer stands out for its ability to capture hierarchical, multi-scale representations, making it a popular backbone for segmentation in videos. However, despite its window-attention scheme, it still incurs a high computational cost, especially in larger variants commonly used for dense prediction in videos. This remains a major bottleneck for real-time, resource-constrained applications. Whilst token reduction methods have been proposed to alleviate this, the window-based attention mechanism of Swin requires a fixed number of tokens per window, limiting the applicability of conventional pruning techniques. Meanwhile, training-free token clustering approaches have shown promise in image segmentation while maintaining window consistency. Nevertheless, they fail to exploit temporal redundancy, missing a key opportunity to further optimize video segmentation performance. We introduce Temporal Cluster Assignment (TCA), a lightweight and effective, fine-tuning-free strategy that enhances token clustering by leveraging temporal coherence across frames. Instead of indiscriminately dropping redundant tokens, TCA refines token clusters using temporal correlations, thereby retaining fine-grained details while significantly reducing computation. Extensive evaluations on YouTube-VIS 2019, YouTube-VIS 2021, OVIS, and a private surgical video dataset show that TCA consistently boosts the accuracy-speed trade-off of existing clustering-based methods. Our results demonstrate that TCA generalizes competently across both natural and domain-specific videos.