Research map / Image generation / Fast diffusion sampling
Anomaly-Preference Image Generation
Fuyun Wang, Yuanzhi Wang, Xu Guo, Sujia Huang, Tong Zhang, Dan Wang, Xin Liu, Hui Yan, Zhen Cui
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 CPO implicit preference alignment (ablation) | APO · MVTec capsule/grid/wood localization | AUC-P / AP-P / F1-P | Improved |
| add Time-Aware Capacity Allocation (TACA) (ablation) | APO · MVTec capsule/grid/wood localization | AUC-P / AP-P / F1-P | Improved |
| set TACA minimum rank kmin to 4 (versus 1, 2, 8, 16) (ablation) | APO · MVTec generation (IS, IC-LPIPS) | IS / IC-LPIPS | Mixed |
| use implicit preference alignment with TACA and hierarchical sampling instead of prior generators | APO · MVTec generation quality | average IS: 2.02 -> 2.14 | Improved |
| train U-Net on APO synthetic anomalies instead of baseline generators | U-Net detector · MVTec AD pixel-level | AUC-P: 99.1 -> 99.6 | Improved |
| train U-Net on APO synthetic anomalies instead of AnomalyDiffusion data | U-Net detector · MVTec AD image-level | AP-I: 99.7 -> 99.5 | No clear effect |
| train ResNet-34 classifier on APO synthetic anomalies | ResNet-34 · MVTec anomaly classification | accuracy (%): 70.25 -> 72.06 | Improved |
| use U-Net trained on APO data instead of supervised detectors | anomaly localization · MVTec AD | AUROC/AP average: 99.0/78.6 -> 99.6/87.7 (vs PRN) | 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.
- AnomalyDiffusion — AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024) · better on 5 of 6 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
- $\alpha$-DPO: Robust Preference Alignment for Diffusion Models via $\alpha$ Divergence (ICLR 2026 · similarity 0.88)
- Diffusion-NPO: Negative Preference Optimization for Better Preference Aligned Generation of Diffusion Models (ICLR 2025 · similarity 0.86)
- Inference-Time Alignment of Diffusion Models with Direct Noise Optimization (ICML 2025 · similarity 0.86)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.85)
- One-to-Normal: Anomaly Personalization for Few-shot Anomaly Detection (NeurIPS 2024 · similarity 0.85)
- The Crystal Ball Hypothesis in diffusion models: Anticipating object positions from initial noise (ICLR 2025 · similarity 0.84)
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.
