Anomaly Classification on MVTec AD (Accuracy Breakdown)
88.37Accuracy (bottle)AnoDiff
Evaluation results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AnoDiffBackbone=ResNet-342026.03 | 88.37 | — | 76.56 | 44 | 58.06 | 60 | 81.25 | 65.08 | 82.81 | 64.58 | 29.63 | 92.98 | 75 | 78.57 | 86.59 | 70.25 | — | |
| O2MAGBackbone=ResNet-342026.03 | 86.05 | — | 76.56 | 77.33 | 74.19 | 80 | 95.83 | 92.06 | 84.38 | 72.92 | 56.79 | 100 | 92.86 | 78.57 | 85.37 | 82.35 | — | |
| SeaSBackbone=ResNet-342026.03 | 81.4 | — | 48.44 | 33.33 | 38.71 | 47.5 | 81.25 | 61.9 | 68.75 | 18.75 | 56.79 | 89.47 | 57.14 | 76.19 | 34.15 | 56.7 | — | |
| TF2Backbone=ResNet-342026.03 | 79.36 | — | 60 | 53.12 | 55.05 | 47.36 | 88.88 | 46.06 | 72.04 | 34.75 | 40.33 | 88.09 | 67.5 | 70 | 63.86 | 61.88 | — | |
| DualAnoDiff†Backbone=ResNet-34, Implementation Source=paper values2025.07 | 79.07 | — | 78.12 | 70.67 | 79.03 | 80 | 89.58 | 90.48 | 89.06 | 56.25 | 70.37 | 100 | 71.43 | 85.71 | 75.61 | 79.67 | — | |
| MAGICBackbone=ResNet-342025.07 | 76.74 | — | 68.75 | 58.67 | 62.9 | 60 | 97.92 | 85.71 | 90.62 | 67.71 | 82.72 | 100 | 89.29 | 73.81 | 78.05 | 78.06 | — | |
| DualAnoDiffBackbone=ResNet-34, Implementation Source=official code2025.07 | 72.09 | — | 56.25 | 48 | 70.97 | 60 | 85.42 | 84.13 | 76.56 | 33.33 | 58.02 | 98.25 | 71.43 | 71.43 | 73.17 | 68.5 | — | |
| DualAnoBackbone=ResNet-342026.03 | 67.44 | — | 57.81 | 50.67 | 62.9 | 67.5 | 79.17 | 84.13 | 73.44 | 41.67 | 39.51 | 100 | 78.57 | 90.48 | 24.39 | 65.55 | — | |
| DualAnoDiff*Backbone=ResNet-34, Implementation Source=pretrained weights2025.07 | 58.14 | — | 56.25 | 53.33 | 67.74 | 70 | 83.33 | 95.24 | 70.31 | 50 | 48.15 | 100 | 67.86 | 92.86 | 26.83 | 67.15 | — | |
| DFMGANBackbone=ResNet-342026.03 | 56.59 | — | 45.31 | 37.23 | 47.31 | 40.83 | 81.94 | 49.73 | 64.58 | 29.52 | 37.45 | 74.85 | 52.38 | 49.21 | 27.64 | 49.61 | — | |
| AnoDiffClassifier=ResNet-342025.07 | — | 64.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnoDiffLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 68.4 | |
| AnoGenClassifier=ResNet-342025.07 | — | 56.92 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AnoGenLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 61.4 | |
| AnyDoorLearning Paradigm=Training-Free Few-shot Image Insertion2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 61.8 | |
| DualAnoDiffClassifier=ResNet-342025.07 | — | 68.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DualAnoDiffLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 64 | |
| InsertAnyLearning Paradigm=Training-Free Few-shot Image Insertion2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 63.6 | |
| MAGICClassifier=ResNet-342025.07 | — | 78.06 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeaSClassifier=ResNet-342025.07 | — | 52.73 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SeaSLearning Paradigm=Few-shot Learning Anomaly Generation2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 59.1 | |
| UniDG-RFTLearning Paradigm=Training-Free Few-shot Image Insertion, Synthesis Strategy=RFT2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 77.4 | |
| UniDG-SFTLearning Paradigm=Training-Free Few-shot Image Insertion, Synthesis Strategy=SFT2026.04 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 74.9 |