Image-level Anomaly Detection on VisA
98.7AUCDinomaly
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Dinomaly2025.10 | 98.7 | 96.2 | 98.9 | — | |
| ShortcutBreaker2025.10 | 98.7 | 95.9 | 98.9 | — | |
| UniADetReference Image=Used, Shots=1, Backbone=DINOv3 (ViT-L/16)2026.01 | 95.8 | — | — | 95.5 | |
| Recontrast2025.10 | 95.5 | 92 | 96.4 | — | |
| UniADetReference Image=Used, Shots=1, Backbone=DINOv2 ViT-L/142026.01 | 95.2 | — | — | 95.6 | |
| MambaAD2025.10 | 94.3 | 89.4 | 94.5 | — | |
| MAGIC2025.07 | 94.28 | 88.12 | 92.8 | — | |
| UniVADsetting=1-normal-shot2024.12 | 93.5 | — | — | — | |
| GroundingAnomalySegmentation Model=Segformer2026.04 | 91.74 | 85.73 | 91.07 | — | |
| SeaSSegmentation Model=Segformer2026.04 | 91.31 | 85.17 | 90.94 | — | |
| DualAnoDiffSegmentation Model=Segformer2026.04 | 90.99 | 82.11 | 90.07 | — | |
| GroundingAnomalySegmentation Model=UPerNet2026.04 | 90.81 | 84.05 | 91.07 | — | |
| AdaptCLIPReference Image=Not Used, Shots=02026.01 | 90.5 | — | — | 92.3 | |
| ViTAD2025.10 | 90.5 | 86.3 | 91.7 | — | |
| SeaSSegmentation Model=UPerNet2026.04 | 90.34 | 84.33 | 90.73 | — | |
| GroundingAnomalySegmentation Model=Average2026.04 | 90.23 | 83.6 | 89.76 | — | |
| AnoGen2025.07 | 89.64 | 81.28 | 85.64 | — | |
| LAKEBackbone=ViT-L/14@336px, Shot Setting=64-shot2026.04 | 89.4 | 86.2 | 90 | — | |
| UniADetReference Image=Used, Shots=1, Backbone=CLIP ViT-L/14@336px2026.01 | 89.3 | — | — | 90.6 | |
| SeaSSegmentation Model=Average2026.04 | 89.09 | 83.33 | 89.44 | — | |
| UniAD2025.10 | 88.8 | 85.8 | 90.8 | — | |
| DualAnoDiffSegmentation Model=UPerNet2026.04 | 88.71 | 81.43 | 89.64 | — | |
| SeaS2025.07 | 88.6 | 81.51 | 87.44 | — | |
| SSVPsetting=zero-shot2026.01 | 88.2 | 85.4 | 90.4 | — | |
| DualAnoDiffSegmentation Model=Average2026.04 | 88.16 | 81.56 | 88.81 | — | |
| GroundingAnomalySegmentation Model=BiSeNet V22026.04 | 88.13 | 81.03 | 87.14 | — | |
| AnomalyGPTsetting=1-normal-shot2024.12 | 87.4 | — | — | — | |
| Bayes-PFLsetting=zero-shot2026.01 | 87.1 | 83.4 | 89 | — | |
| Bayes-PFLBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 87 | 84.1 | 89.2 | — | |
| Bayes-PFLsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 86.8 | — | 89.3 | — | |
| DiAD2025.10 | 86.8 | 85.1 | 88.3 | — | |
| DualAnoDiff2025.07 | 86.45 | 75.47 | 82.05 | — | |
| AnomalyCLIP+Reference Image=Not Used, Shots=02026.01 | 86.1 | — | — | 87.7 | |
| SeaSSegmentation Model=BiSeNet V22026.04 | 85.61 | 80.49 | 86.64 | — | |
| AnoDiff2025.07 | 85.03 | 78.05 | 81.24 | — | |
| AnomalyDiffusionSegmentation Model=Segformer2026.04 | 84.77 | 79.54 | 84.53 | — | |
| DualAnoDiffSegmentation Model=BiSeNet V22026.04 | 84.77 | 81.13 | 86.72 | — | |
| VisualAD(CLIP)Backbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 84.7 | 82.5 | 87.6 | — | |
| VisualAD (CLIP)Backbone=CLIP2026.04 | 84.7 | 82.5 | 87.6 | — | |
| WINCLIPsetting=1-normal-shot2024.12 | 83.8 | — | — | — | |
| MoECLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 83.6 | — | 86.2 | — | |
| InCtrlReference Image=Not Used, Shots=02026.01 | 83.2 | — | — | 84.1 | |
| AnomalyDiffusionSegmentation Model=UPerNet2026.04 | 83.18 | 78.88 | 84.08 | — | |
| VisualAD(DINOv2)Backbone=ViT-L/14, Evaluation Protocol=Zero-shot2026.03 | 83.1 | 81.4 | 86.8 | — | |
| VisualAD (DINOv2)Backbone=DINOv22026.04 | 83.1 | 81.4 | 86.8 | — | |
| AdaCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 83 | 81.6 | 84.9 | — | |
| AdaCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial, re-trained status=True2026.03 | 83 | — | 85.5 | — | |
| AdaCLIPsetting=zero-shot2026.01 | 82.6 | 81.3 | 84.4 | — | |
| AnomalyCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 82.1 | 80.4 | 85.4 | — | |
| AnomalyCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 82.1 | — | 85.4 | — | |
| AnomalyCLIPsetting=zero-shot2026.01 | 82 | 80.1 | 85.2 | — | |
| AnomalyDiffusionSegmentation Model=Average2026.04 | 81.35 | 77.18 | 82.12 | — | |
| AnomalyCLIPBackbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 81 | 80.3 | 84.4 | — | |
| AnomalyCLIP2026.04 | 81 | 80.3 | 84.4 | — | |
| WinCLIP+Reference Image=Not Used, Shots=02026.01 | 80 | — | — | 81.7 | |
| CLIP-ADBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 79.8 | 79.2 | 84.3 | — | |
| AdaCLIPBackbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 79.7 | 79.63 | 83.2 | — | |
| AdaCLIP2026.04 | 79.7 | 79.6 | 83.2 | — | |
| AA-CLIPsetting=Zero-shot setting (ZSAD), domain=Industrial, re-trained status=True2026.03 | 79.2 | — | 83.7 | — | |
| WinCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 78.1 | 79 | 77.5 | — | |
| WinCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 78.1 | — | 77.5 | — | |
| APRIL-GANBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 78 | 78.7 | 81.4 | — | |
| APRIL-GANsetting=zero-shot2026.01 | 77.9 | 78.4 | 81.1 | — | |
| WinCLIPsetting=zero-shot2026.01 | 77.6 | 78.8 | 77.1 | — | |
| April-GANsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 77.5 | — | 80.8 | — | |
| APRIL-GANBackbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 77.4 | 78.6 | 80.9 | — | |
| AnomalyDiffusionSegmentation Model=BiSeNet V22026.04 | 76.11 | 73.13 | 77.74 | — | |
| WinCLIPBackbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 75.6 | 78.2 | 78.8 | — | |
| WinCLIP2026.04 | 75.6 | 78.2 | 78.8 | — | |
| PatchCoresetting=1-normal-shot2024.12 | 74.8 | — | — | — | |
| DFMGANSegmentation Model=Segformer2026.04 | 72.12 | 71.46 | 71.93 | — | |
| DFMGANSegmentation Model=UPerNet2026.04 | 71.69 | 70.7 | 71.64 | — | |
| MedCLIPsetting=1-normal-shot2024.12 | 69 | — | — | — | |
| DFMGANSegmentation Model=Average2026.04 | 68.96 | 69.55 | 68.73 | — | |
| CLIP-ADBackbone=ViT-L/14@336px, Evaluation Protocol=Zero-shot2026.03 | 66.2 | 74.3 | 71.4 | — | |
| CLIP-AD2026.04 | 66.2 | 74.3 | 71.4 | — | |
| DFMGANSegmentation Model=BiSeNet V22026.04 | 63.07 | 66.48 | 62.63 | — | |
| UniADsetting=1-normal-shot2024.12 | 61.3 | — | — | — | |
| OpenCLIPBackbone=ViT-L/142026.04 | 60.2 | 73 | 66.2 | — | |
| ComADsetting=1-normal-shot2024.12 | 53.9 | — | — | — |