The classifier judged this item not relevant to AI security, so it is left out of lists, feeds, the API and dataset releases. If that is wrong, report a correction on this page.
Reconstructive Visual Tuning for Weakly Supervised Video Anomaly Detection
- Published
- Record updated
Summary
This paper proposes Reconstructive Visual Tuning (RVT) for weakly supervised video anomaly detection (WS-VAD), which must find anomaly frames using only video-level labels. Prior methods adapting CLIP receive only coarse supervision and miss fine-grained anomaly patterns. RVT regularizes visual representations by reconstructing input video features rather than raw RGB values, and adds a global-local collaborative temporal modeling module (GLCT) to capture temporal dependencies. The authors report consistent improvements across several VAD datasets and competitive results against state-of-the-art methods.