Pixel-level Anomaly Detection on VisA
98.5AUROCPatchCore
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PatchCoreCategory=Obj & texture, |C|=12, Evaluation Protocol=Full-shot2023.10 | 98.5 | 92.2 | — | — | |
| RD4ADCategory=Obj & texture, |C|=12, Evaluation Protocol=Full-shot2023.10 | 98.4 | 91.2 | — | — | |
| UniADetReference Image=Used, Shots=1, Backbone=DINOv2 ViT-L/142026.01 | 98.2 | — | — | 42.1 | |
| UniADetReference Image=Used, Shots=1, Backbone=DINOv3 (ViT-L/16)2026.01 | 98.2 | — | — | 44 | |
| UniADetReference Image=Used, Shots=1, Backbone=CLIP ViT-L/14@336px2026.01 | 97.5 | — | — | 32.8 | |
| AdaptCLIPReference Image=Not Used, Shots=02026.01 | 96.8 | — | — | 38.9 | |
| AnomalyCLIP+Reference Image=Not Used, Shots=02026.01 | 96.4 | — | — | 24.8 | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=DINOv3 ViT-L/162026.01 | 96.3 | — | — | 32.5 | |
| SSVPsetting=zero-shot2026.01 | 96.2 | 90.5 | 30.1 | — | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=DINOv2-R ViT-L/162026.01 | 95.9 | — | — | 32.7 | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=CLIP ViT-L/14@336px2026.01 | 95.8 | — | — | 28 | |
| AdaCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 95.7 | — | — | 31 | |
| AdaptCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 95.7 | — | — | 26.1 | |
| FiLoProtocol=Zero-shot2026.05 | 95.7 | 80.6 | — | — | |
| FAPromptProtocol=Zero-shot2026.05 | 95.7 | 86.8 | — | — | |
| Bayes-PFLBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 95.6 | 88.9 | 29.8 | — | |
| Bayes-PFLEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 95.6 | — | — | 29.8 | |
| MoECLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 95.6 | — | 26.1 | — | |
| AnomalyCLIPCategory=Obj & texture, |C|=12, Evaluation Protocol=Zero-shot2023.10 | 95.5 | 87 | — | — | |
| AnomalyCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 95.5 | 87 | 21.3 | — | |
| AnomalyCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 95.5 | — | — | 21.3 | |
| AnomalyCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 95.5 | — | 21.2 | — | |
| AdaCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial, re-trained status=True2026.03 | 95.5 | — | 28.6 | — | |
| Bayes-PFLsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 95.5 | — | 29.2 | — | |
| AnomalyCLIPsetting=zero-shot2026.01 | 95.2 | 86.4 | 21 | — | |
| AdaCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 95.1 | 71.3 | 29.2 | — | |
| CLIP-ADBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 95 | 86.9 | 26.3 | — | |
| MRADProtocol=Zero-shot2026.05 | 94.9 | 84.9 | — | — | |
| AA-CLIPProtocol=Zero-shot2026.05 | 94.8 | 81.4 | — | — | |
| EntroADProtocol=Zero-shot2026.05 | 94.8 | 87.4 | — | — | |
| AdaCLIPsetting=zero-shot2026.01 | 94.7 | 70.8 | 26.4 | — | |
| WinCLIP+Reference Image=Not Used, Shots=02026.01 | 94.7 | — | — | 15.8 | |
| AA-CLIPsetting=Zero-shot setting (ZSAD), domain=Industrial, re-trained status=True2026.03 | 94.7 | — | 24.2 | — | |
| AnomalyCLIPProtocol=Zero-shot2026.05 | 94.4 | 86.7 | — | — | |
| APRIL-GANBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 94.2 | 86.8 | 25.7 | — | |
| April-GANsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 94.2 | — | 25.8 | — | |
| APRIL-GANsetting=zero-shot2026.01 | 94.1 | 86.4 | 25.2 | — | |
| Bayes-PFLsetting=zero-shot2026.01 | 93.2 | 89.2 | 29.7 | — | |
| InCtrlReference Image=Not Used, Shots=02026.01 | 89 | — | — | 17.7 | |
| WinCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 79.6 | 56.8 | 5 | — | |
| WinCLIPsetting=Zero-shot setting (ZSAD), domain=Industrial2026.03 | 79.6 | — | 5 | — | |
| WinCLIPsetting=zero-shot2026.01 | 79.3 | 56.2 | 4.9 | — | |
| WinCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 73.2 | — | — | 5.4 | |
| WinCLIPProtocol=Zero-shot2026.05 | 73.2 | 51 | — | — |