Research map / Label-efficient and robust learning / OOD and anomaly detection
Quality-Aware Language-Conditioned Local Auto-Regressive Anomaly Synthesis and Detection
Long Qian, Bingke Zhu, Yingying Chen, Ming Tang, Jinqiao 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 |
|---|---|---|---|
| replace DTD/Perlin synthesis with ARAS (ablation) | DRAEM · MVTec AD | image/pixel AUROC: 98.0/97.2 -> 98.6/97.9 | Improved |
| replace SDAS diffusion synthesis with ARAS (ablation) | RealNet · MVTec AD | pixel AUROC: 99.0 -> 99.3 | Improved |
| replace SDAS diffusion synthesis with ARAS (ablation) | RealNet · MVTec AD | image AUROC: 99.6 -> 99.6 | No clear effect |
| add hinge quality-aware weighting (ablation) | RealNet+ARAS · MVTec AD | pixel AUROC: 99.3 -> 99.8 | Improved |
| add softmax quality-aware weighting (QAW) (ablation) | RealNet+ARAS · MVTec AD | pixel AUROC: 99.3 -> 99.9 | Improved |
| use softmax rather than hinge weighting (ablation) | RealNet+ARAS · MVTec AD | pixel AUROC: 99.8 -> 99.9 | Improved |
| QARAD over RealNet | RealNet · VisA | image/pixel AUROC: 97.8/98.8 -> 98.9/99.8 | Improved |
| QARAD over RealNet | RealNet · MVTec AD | image/pixel AUROC: 99.6/99.0 -> 99.7/99.8 | Improved |
| QARAD over RealNet | RealNet · BTAD | image/pixel AUROC: 96.1/97.9 -> 96.7/98.0 | Improved |
| ARAS in place of diffusion-based synthesis | RealNet · synthesis time per image | seconds: 7.51 -> 1.49 | Improved |
| add QAW (CLIP similarity scoring) | QARAD · detector training | train time (s/ep): 100.89 -> 110.36 | Hurt |
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.
- RealNet — RealNet: A Feature Selection Network with Realistic Synthetic Anomaly for Anomaly Detection (CVPR 2024) · better on 6 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
- AnoStyler: Text-Driven Localized Anomaly Generation via Lightweight Style Transfer (AAAI 2026 · similarity 0.89)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.86)
- Generating and Reweighting Dense Contrastive Patterns for Unsupervised Anomaly Detection (AAAI 2024 · similarity 0.86)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.86)
- A Diffusion-Based Framework for Multi-Class Anomaly Detection (AAAI 2024 · similarity 0.86)
- Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · 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.
