Anomaly Synthesis Quality on MVTec-AD
0.37ILDualAnoDiff
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DualAnoDiffLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | 0.37 | 0.18 | 3.32 | 3.57 | 3.67 | 3.55 | |
| SeaSLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | 0.34 | 0.19 | 3.26 | 3.76 | 3.74 | 3.58 | |
| AnyDoorLearning Paradigm=Training-Free Few-shot Image Insertion2026.04 | 0.33 | 0.14 | 3.48 | 3.76 | 3.83 | 3.7 | |
| UniDG-SFTLearning Paradigm=Training-Free Few-shot Image Insertion, Synthesis Strategy=SFT2026.04 | 0.29 | 0.18 | 3.9 | 4.15 | 4.18 | 4.09 | |
| AnoDiffLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | 0.28 | 0.14 | 3.84 | 3.97 | 4.18 | 4.01 | |
| UniDG-RFTLearning Paradigm=Training-Free Few-shot Image Insertion, Synthesis Strategy=RFT2026.04 | 0.28 | 0.16 | 4.04 | 4.21 | 4.2 | 4.24 | |
| AnoGenLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | 0.27 | 0.11 | 3.5 | 3.87 | 3.91 | 3.77 | |
| InsertAnyLearning Paradigm=Training-Free Few-shot Image Insertion2026.04 | 0.26 | 0.13 | 3.81 | 3.98 | 4.12 | 3.99 |