Pixel-level Anomaly Detection on CVC-ClinicDB
90.7AUROCAVA-DINO
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| AVA-DINOProtocol=Zero-shot protocol, Reference images=Not available, Training dataset=ViSA2026.05 | 90.7 | — | — | 57.2 | |
| Bayes-PFLBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 89.6 | 78.4 | 53.2 | — | |
| APRIL-GANBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 86 | 41.2 | 38.8 | — | |
| AnomalyCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 85.9 | 69.6 | 42.2 | — | |
| CLIP-ADBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 85.6 | 57 | 51.7 | — | |
| AdaCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 84.3 | 14.6 | 36 | — | |
| AnomalyCLIPProtocol=Zero-shot protocol, Reference images=Not available, Training dataset=ViSA2026.05 | 82.9 | — | — | 42.1 | |
| AdaCLIPProtocol=Zero-shot protocol, Reference images=Not available, Training dataset=ViSA2026.05 | 82.8 | — | — | 40.9 | |
| WinCLIPBackbone=ViT-L-14-336, Resolution=518x5182025.03 | 70.7 | 32.5 | 19.4 | — | |
| Bayes-PFLProtocol=Zero-shot protocol, Reference images=Not available, Training dataset=ViSA2026.05 | 70.7 | — | — | 38.4 | |
| WinCLIPProtocol=Zero-shot protocol, Reference images=Not available, Training dataset=ViSA2026.05 | 51.2 | — | — | 24.4 |