Research map / Video / Video segmentation and tracking
Language-guided Open-world Video Anomaly Detection under Weak Supervision
Zihao Liu, Xiaoyu Wu, Jianqin Wu, Xuxu Wang, Linlin Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add dynamic video synthesis loss (ablation) | LaGoVAD · zero-shot detection avg (7 datasets) | AUC/AP avg: 73.51 -> 76.42 | Improved |
| add hard-negative contrastive loss (ablation) | LaGoVAD · zero-shot detection avg (7 datasets) | AUC/AP avg: 73.96 -> 76.42 | Improved |
| add language guidance (ablation) | LaGoVAD · zero-shot classification avg | accuracy/F1 avg: 46.23 -> 52.57 | Improved |
| add language guidance (ablation) | LaGoVAD · PreVAD val | detection score: 69.87 -> 69.98 | No clear effect |
| remove dynamic synthesis loss (ablation) | LaGoVAD · PreVAD val | detection score: 69.98 -> 65.73 | Hurt |
| train VadCLIP on PreVAD instead of UCF-Crime/XD | VadCLIP · UCF-Crime classification | accuracy: 38.28 -> 45.52 | Improved |
| train VadCLIP on PreVAD instead of UCF | VadCLIP · XD-Violence classification | accuracy: 46.38 -> 71.38 | Improved |
| replace VadCLIP architecture with LaGoVAD, both on PreVAD | LaGoVAD · XD-Violence classification | accuracy: 71.38 -> 78.13 | Improved |
| replace VadCLIP with LaGoVAD | LaGoVAD · UCF-Crime classification | F1: 17.81 -> 16.64 | Hurt |
| language-guided definition conditioning vs fixed definition under concept drift | LaGoVAD · XD-drift@5 | AUC: 81.7 -> 85.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HolmesVAU — Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · similarity 0.93)
- Harnessing Large Language Models for Training-free Video Anomaly Detection (CVPR 2024 · similarity 0.91)
- MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection (NeurIPS 2025 · similarity 0.90)
- Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025 · similarity 0.90)
- HAWK: Learning to Understand Open-World Video Anomalies (NeurIPS 2024 · similarity 0.90)
- Advancing Video Anomaly Detection: A Concise Review and a New Dataset (NeurIPS 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
