Action Typicality and Uniqueness Learning for Zero-Shot Video Anomaly Detection

Canhui Tang, Sanping Zhou, Haoyue Shi, Hao Sun, Le Wang

Keywords: video anomaly detection, zero-shot, skeleton-based

TL;DR: A zero-shot video anomaly detection framework from the perspective of action typicality and uniqueness.

Abstract: Zero-Shot Video Anomaly Detection (ZS-VAD) is an urgent task in scenarios where the target video domain lacks training data due to various concerns, \emph{e.g.}, data privacy. The skeleton-based approach is a promising way to achieve ZS-VAD as it eliminates domain disparities in both background and human appearance. However, existing methods only learn low-level skeleton representation and rely on the domain-specific normality boundary, which cannot generalize well to new scenes with different normal and abnormal behavior patterns. In this paper, we propose a novel skeleton-based zero-shot video anomaly detection framework, which captures both scene-generic typical anomalies and scene-adaptive unique anomalies. Firstly, we introduce a language-guided typicality modeling module that projects skeleton snippets into action semantic space and learns generalizable typical distributions of normal and abnormal behavior. Secondly, we propose a test-time context uniqueness analysis module to finely analyze the spatio-temporal differences between skeleton snippets and then derive scene-adaptive boundaries. Without using any training samples from the target domain, our method achieves state-of-the-art results on four large-scale VAD datasets: ShanghaiTech, UBnormal, NWPU, and UCF-Crime. The Code will be publicly available.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3575