Research map / Video / Video segmentation and tracking
Towards Open Set Video Anomaly Detection
Yuansheng Zhu, Wentao Bao, Qi Yu
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 triplet loss (ablation) | OpenVAD · XD-Violence (1 seen anomaly classes) | AUC-PR: 45.26 -> 45.65 | Improved |
| add triplet loss (ablation) | OpenVAD · XD-Violence (1 seen anomaly classes) | AUC-ROC: 61.30 -> 72.50 | Improved |
| add triplet loss (ablation) | OpenVAD · XD-Violence (4 seen anomaly classes) | AUC-PR: 69.14 -> 69.61 | Improved |
| add evidence criteria for instance selection (ablation) | OpenVAD without triplet loss · XD-Violence (1 seen anomaly classes) | AUC-PR: 41.79 -> 45.26 | Improved |
| add evidence criteria for instance selection (ablation) | OpenVAD without triplet loss · XD-Violence (1 seen anomaly classes) | AUC-ROC: 67.07 -> 61.30 | Hurt |
| add evidence criteria for instance selection (ablation) | OpenVAD without triplet loss · XD-Violence (4 seen anomaly classes) | AUC-PR: 68.82 -> 69.14 | Improved |
| add flow-generated pseudo anomalies (ablation) | OpenVAD without other components · XD-Violence (1 seen anomaly classes) | AUC-PR: 40.94 -> 41.79 | Improved |
| add flow-generated pseudo anomalies (ablation) | OpenVAD without other components · XD-Violence (4 seen anomaly classes) | AUC-PR: 66.61 -> 68.82 | Improved |
| replace two FC layers with GCNs as feature encoder (ablation) | OpenVAD · XD-Violence (closed set, all anomaly classes) | AUC-PR: 69.39 -> 77.91 | Improved |
| use evidence-based selection instead of top-k (ablation) | OpenVAD · XD-Violence (closed set, all anomaly classes) | AUC-PR: 77.43 -> 77.91 | Improved |
| add triplet loss to NF density scoring (ablation) | normalizing flow scorer · XD-Violence (closed set, all anomaly classes) | AUC-PR: 52.10 -> 73.13 | Improved |
| use NFs to generate pseudo anomalies instead of scoring samples by density (ablation) | normalizing flow · XD-Violence (closed set, all anomaly classes) | AUC-PR: 73.13 -> 77.91 | Improved |
Closest papers by meaning
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · similarity 0.90)
- VADTree: Explainable Training-Free Video Anomaly Detection via Hierarchical Granularity-Aware Tree (NeurIPS 2025 · similarity 0.87)
- Prompt-Enhanced Multiple Instance Learning for Weakly Supervised Video Anomaly Detection (CVPR 2024 · similarity 0.87)
- Learning Event Completeness for Weakly Supervised Video Anomaly Detection (ICML 2025 · similarity 0.87)
- Towards Multi-Domain Learning for Generalizable Video Anomaly Detection (NeurIPS 2024 · similarity 0.87)
- Harnessing Large Language Models for Training-free Video Anomaly Detection (CVPR 2024 · similarity 0.86)
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.
