Anomaly Detection on VisA 10% training contamination (test) benchmark leaderboard · SOTA2 Research

Anomaly Detection on VisA 10% training contamination (test)

0.91Average Image-level AUROCFolded(3/4) PatchCore

Evaluation Results

Method
Folded(3/4) PatchCore2026.02	0.91	—	—	—	—	—	—	—	—	—	—	—	—	0.951	0.879
FUN-AD2026.02	0.909	—	—	—	—	—	—	—	—	—	—	—	—	0.929	0.757
Multi-Cue Anomaly Detection framework2026.01	0.908	0.785	0.972	0.971	0.977	0.951	0.737	0.979	0.973	0.901	0.889	0.82	0.943	—	—
SoftPatch2026.02	0.893	—	—	—	—	—	—	—	—	—	—	—	—	0.887	0.797
ADL2026.01	0.845	0.752	0.913	0.909	0.928	0.808	0.609	0.934	0.917	0.821	0.824	0.814	0.912	—	—
DestSeg2026.01	0.83	0.661	0.917	0.85	0.849	0.746	0.652	0.835	0.925	0.851	0.791	0.922	0.964	—	—
LOE2026.01	0.798	0.625	0.866	0.941	0.961	0.671	0.513	0.756	0.813	0.817	0.849	0.798	0.967	—	—
InReaCh2026.02	0.73	—	—	—	—	—	—	—	—	—	—	—	—	0.944	0.705
PatchCore2026.01	0.717	0.551	0.749	0.772	0.717	0.691	0.589	0.769	0.759	0.735	0.715	0.665	0.891	—	—
PaDiM2026.01	0.711	0.29	0.856	0.856	0.836	0.731	0.119	0.723	0.846	0.832	0.769	0.746	0.93	—	—
DRÆM2026.01	0.648	0.465	0.819	0.433	0.809	0.861	0.724	0.749	0.957	0.211	0.235	0.707	0.805	—	—