Learn the interactions: Weakly supervised video anomaly detection with human-object interactions
Maowen Zhou, Erma Rahayu Mohd Faizal Abdullah, Aznul Qalid Md Sabri, Susanto Rahardja · 2026
Read Learn the interactions: Weakly supervised video anomaly detection with human-object interactions by Maowen Zhou, Erma Rahayu Mohd Faizal Abdullah, Aznul Qalid Md Sabri, Susanto Rahardja (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0358538.txt.
Description
Maowen Zhou and three colleagues published a computer vision study in PLOS ONE in 2026. They tackled the task of spotting odd or unsafe events in video when only weak labels are at hand, rather than costly frame-by-frame notes. Their approach leans on how people interact with objects in a scene.
Zhou's system learns from those human-object interactions to judge which clips look normal and which do not. Unusual handling of an object can flag an event that a watcher should review. Using weak labels cuts training cost a great deal. The method suits security and safety settings where staff cannot label hours of video by hand but still need alerts they can trust.
Catalog record
| CREATOR | Maowen Zhou, Erma Rahayu Mohd Faizal Abdullah, Aznul Qalid Md Sabri, Susanto Rahardja |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60947 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |