Research map / Video / Video-language understanding
TargetVAU: Multimodal Anomaly-Aware Reasoning for Target Behavior Understanding in Videos
Lingru Zhou, Peng Wu, Manqing Zhang, Qingsheng Wang, Guansong Pang, Peng Wang
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 Anomaly-focused Temporal Sampler (vs uniform sampling) (ablation) | TargetVAU · UCCD | CIDEr: 0.591 -> 0.665 | Improved |
| add Anomaly-focused Temporal Sampler (ablation) | TargetVAU · UCCD | BLEU: 0.509 -> 0.582 | Improved |
| add frozen Q-Former fusion (vs linear projection) (ablation) | TargetVAU · UCCD | CIDEr: 0.603 -> 0.665 | Improved |
| add Q-Former fusion (ablation) | TargetVAU · UCCD | BLEU: 0.521 -> 0.582 | Improved |
| add Spatio-Temporal Interaction Graph (ablation) | TargetVAU · UCCD | BLEU/CIDEr | Improved |
| add global features to human-centric features (ablation) | TargetVAU · UCCD | BLEU/CIDEr | Improved |
| raise anomaly detection loss weight (2:1:1 vs 1:1:2) (ablation) | TargetVAU · UCCD | CIDEr: 0.665 -> 0.641 | Hurt |
| raise contrastive loss weight (1:2:1 vs 1:1:2) (ablation) | TargetVAU · UCCD | CIDEr: 0.665 -> 0.649 | Hurt |
| raise caption loss weight to 2 (vs equal weights) (ablation) | TargetVAU · UCCD | CIDEr: 0.658 -> 0.665 | Improved |
| TargetVAU vs Holmes-VAU zero-shot | TargetVAU · HIVAU-70K | CIDEr: 1.437 -> 1.424 | Hurt |
| TargetVAU vs Holmes-VAU zero-shot | TargetVAU · HIVAU-70K | ParaScore: 0.909 -> 0.901 | Hurt |
| TargetVAU vs Holmes-VAU for detection | TargetVAU · UCF-Crime | AUC (%): 88.96 -> 89.57 | 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.
- Holmes-VAU — Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025) · better on 12 of 14 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
- Vad-R1: Towards Video Anomaly Reasoning via Perception-to-Cognition Chain-of-Thought (NeurIPS 2025 · similarity 0.90)
- VERA: Explainable Video Anomaly Detection via Verbalized Learning of Vision-Language Models (CVPR 2025 · similarity 0.89)
- Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025 · similarity 0.89)
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · similarity 0.88)
- A Unified Reasoning Framework for Holistic Zero-Shot Video Anomaly Analysis (NeurIPS 2025 · similarity 0.88)
- HAWK: Learning to Understand Open-World Video Anomalies (NeurIPS 2024 · similarity 0.88)
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.
