Research map / Label-efficient and robust learning / OOD and anomaly detection
One-to-More: High-Fidelity Training-Free Anomaly Generation with Attention Control
Haoxiang Rao, Zhao Wang, Chenyang Si, Yan Lyu, Yuanyi Duan, Fang Zhao, Caifeng Shan
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 Dual-Attention Enhancement to TriAG (ablation) | O2MAG · MVTec-AD pixel-level | AP: 81.9 -> 83.3 | Improved |
| add Anomaly-Guided Optimization to TriAG (ablation) | O2MAG · MVTec-AD pixel-level | AP: 81.9 -> 82.7 | Improved |
| add both DAE and AGO to TriAG (ablation) | O2MAG · MVTec-AD pixel-level | AP: 81.9 -> 86.3 | Improved |
| add DAE to TriAG (ablation) | O2MAG · MVTec-AD pixel-level | AUROC: 99.0 -> 99.0 | No clear effect |
| use O2MAG instead of AnomalyDiffusion for synthetic training data | classification with ResNet-34 · MVTec-AD | Accuracy (%): 70.25 -> 82.35 | Improved |
| use O2MAG instead of AnomalyAny (training-free) | U-Net detection · MVTec-AD | F1-P: 65.1 -> 80.6 | Improved |
| use O2MAG instead of baselines for synthetic training data | U-Net detection · MVTec-AD | AP-I: 99.7 -> 99.8 | No clear effect |
| use O2MAG instead of DualAnoDiff for diversity | generation · MVTec-AD | IC-LPIPS: 0.38 -> 0.30 | Hurt |
| use O2MAG instead of baselines for fidelity | generation · MVTec-AD | KID: 62.85 (DFMGAN) -> 45.55 | Improved |
| zero-shot cross-class reference (wood-hole to hazelnut-hole) instead of same-class reference | O2MAG · hazelnut-hole | AP-P: 90.4 -> 86.8 | Hurt |
Closest papers by meaning
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.92)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.90)
- Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation (CVPR 2025 · similarity 0.90)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.89)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.88)
- One-to-Normal: Anomaly Personalization for Few-shot Anomaly Detection (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.
