Reproduce “GlanceVAD: Exploring Glance Supervision for Label-efficient Video Anomaly Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Huaxin Zhang, Xiang Wang, Xiaohao Xu, Xiaonan Huang, Chuchu Han, Yuehuan Wang, Changxin Gao, Shanjun Zhang and 1 more
This work introduces "glance supervision," a new labeling paradigm for video anomaly detection that involves marking a single frame within each anomalous event. The GlanceVAD method leverages this minimal temporal information through Temporal Gaussian Splatting to achieve state-of-the-art performance on UCF-Crime and XD-Violence datasets, improving AUC by 4.99% and AP by 6.00% respectively, while significantly reducing annotation costs compared to fully-supervised approaches.
Key insights
- Glance supervision offers a favorable trade-off, providing sufficient temporal guidance to surpass weak supervision accuracy while maintaining comparable annotation costs.
- Temporal Gaussian Splatting effectively converts sparse glance annotations into continuous anomaly scores, addressing the challenge of limited direct supervision.
- GlanceVAD demonstrates notable robustness to inaccuracies in the precise placement of glance annotations, simplifying the labeling process.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.