Research map / Label-efficient and robust learning / OOD and anomaly detection
Exploiting Completeness and Uncertainty of Pseudo Labels for Weakly Supervised Video Anomaly Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 a multi-head classifier with diversity loss to generate more complete pseudo labels (ablation) | MIL pseudo-label baseline · UCF-Crime | AUC(%): +2.03% | Helped |
| Add iterative uncertainty-aware pseudo-label refinement (ablation) | MIL pseudo-label baseline · UCF-Crime | AUC(%): +1.83% | Helped |
| Add both completeness and uncertainty components (ablation) | MIL pseudo-label baseline · UCF-Crime | AUC(%): 82.86 -> 86.22 | Helped |
| Use 2 classifier heads instead of 1 (ablation) | Completeness enhanced pseudo-label generator · UCF-Crime | AUC(%): 84.69 -> 86.22 | Helped |
| Increase the number of heads from 2 to 3, 4, or 5 (ablation) | Completeness enhanced pseudo-label generator · UCF-Crime | AUC(%): 86.22; 85.05; 85.90; 85.10 | Mixed |
| Set diversity loss weight α to 100 rather than 10 (ablation) | Completeness enhanced pseudo-label generator · UCF-Crime | AUC(%): 86.22 -> 75.06 | Hurt |
| Vary diversity loss weight α from 0.01 through 10 (ablation) | Completeness enhanced pseudo-label generator · UCF-Crime | AUC(%): 84.56; 84.43; 84.82; 86.22 | Mixed |
| Select reliable clips using uncertainty instead of randomly selecting samples (ablation) | Self-training · UCF-Crime | AUC | Helped |
| Use I3D+VGGish features instead of I3D features | Ours · XD-Violence | AP(%): 78.74 -> 81.43 | Helped |
| Use the proposed method instead of WSAL | UCF-Crime · UCF-Crime | AUC(%): 0.92% | Helped |
| Use the proposed method instead of MIST | TAD · TAD | AUC(%): 2.40% | Helped |
| Use the proposed method instead of RTFM | TAD · TAD | AUC(%): 2.02% | Helped |
Closest papers by meaning
- Learning Event Completeness for Weakly Supervised Video Anomaly Detection (ICML 2025 · similarity 0.89)
- TLMA: Mitigating the Impact of Weakly Labeled Information for Video Anomaly Detection (CVPR 2026 · similarity 0.89)
- MIST: Multiple Instance Self-Training Framework for Video Anomaly Detection (CVPR 2021 · similarity 0.89)
- Text Prompt with Normality Guidance for Weakly Supervised Video Anomaly Detection (CVPR 2024 · similarity 0.89)
- Prompt-Enhanced Multiple Instance Learning for Weakly Supervised Video Anomaly Detection (CVPR 2024 · similarity 0.86)
- FrameShield: Adversarially Robust Video Anomaly Detection (NeurIPS 2025 · similarity 0.85)
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.