Image-level Anomaly Detection on MVTec
97.6Average ROCAUCUniADet
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UniADetReference Image=Used, Shots=1, Backbone=DINOv2 ViT-L/142026.01 | 97.6 | 98.6 | |
| UniADetReference Image=Used, Shots=1, Backbone=DINOv3 (ViT-L/16)2026.01 | 97.2 | 98.1 | |
| UniADetReference Image=Used, Shots=1, Backbone=CLIP ViT-L/14@336px2026.01 | 96.3 | 98.1 | |
| AnomalyCLIP+Reference Image=Not Used, Shots=02026.01 | 95.2 | 97.2 | |
| AdaptCLIPReference Image=Not Used, Shots=02026.01 | 94.5 | 97.5 | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=DINOv3 ViT-L/162026.01 | 94 | 97.1 | |
| WinCLIP+Reference Image=Not Used, Shots=02026.01 | 93.6 | 96.8 | |
| AdaptCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 93.5 | 96.7 | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=DINOv2-R ViT-L/162026.01 | 93.5 | 97.2 | |
| UniADetEvaluation Protocol=Zero-shot, Language-free=true, Backbone=CLIP ViT-L/14@336px2026.01 | 92.4 | 96.1 | |
| Bayes-PFLEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 92.3 | 96.7 | |
| AnomalyCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 91.6 | 96.4 | |
| InCtrlReference Image=Not Used, Shots=02026.01 | 91.3 | 95.2 | |
| AdaCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 90.7 | 95.6 | |
| WinCLIPEvaluation Protocol=Zero-shot, Language-free=false, Backbone=CLIP ViT-L/14@336px2026.01 | 90.4 | 95.6 | |
| SPADE2020.05 | 85.5 | — | |
| ITAE2020.05 | 83.9 | — | |
| GANomaly2020.05 | 76.2 | — | |
| AE_L22020.05 | 75.4 | — | |
| Geom2020.05 | 67.2 | — |