AutoQ-VIS: Improving Unsupervised Video Instance Segmentation via Automatic Quality Assessment

TL;DR

AutoQ-VIS introduces an unsupervised video instance segmentation framework that uses automatic quality assessment to improve pseudo-labels, enabling effective domain adaptation from synthetic to real videos without human annotations.

Contribution

It proposes a novel quality-guided self-training approach that bridges the synthetic-to-real domain gap in unsupervised VIS, achieving state-of-the-art results.

Findings

Achieves 52.6% AP50 on YouTubeVIS-2019 val set.

Surpasses previous state-of-the-art VideoCutLER by 4.4%.

Requires no human annotations for training.

Abstract

Video Instance Segmentation (VIS) faces significant annotation challenges due to its dual requirements of pixel-level masks and temporal consistency labels. While recent unsupervised methods like VideoCutLER eliminate optical flow dependencies through synthetic data, they remain constrained by the synthetic-to-real domain gap. We present AutoQ-VIS, a novel unsupervised framework that bridges this gap through quality-guided self-training. Our approach establishes a closed-loop system between pseudo-label generation and automatic quality assessment, enabling progressive adaptation from synthetic to real videos. Experiments demonstrate state-of-the-art performance with 52.6 $\text{AP}_{50}$ on YouTubeVIS-2019 val set, surpassing the previous state-of-the-art VideoCutLER by 4.4$\%$, while requiring no human annotations. This demonstrates the viability of quality-aware self-training for unsupervised VIS. The source code of our method is available at https://github.com/wcbup/AutoQ-VIS.