INP-Former ViT-B (model-unified multi-class)

Reported on 5 benchmarks across 1 task · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Methodology5 results

Anomaly DetectiononVisA
Detection AUROC· 2025-03-04
98.9
best: 99.8 (UniNet)
Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection arXiv:2503.02424
Anomaly DetectiononVisA
F1-Score· 2025-03-04
96.6
best: 98.3 (GLAD)
Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection arXiv:2503.02424
Anomaly DetectiononVisA
Segmentation AUPRO· 2025-03-04
94.4
best: 96 (DiffusionAD)
Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection arXiv:2503.02424
Anomaly DetectiononVisA
Segmentation AUPRO (until 30% FPR)· 2025-03-04
94.4
best: 96.1 (AnomalyDINO-S (full-shot))
Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection arXiv:2503.02424
Anomaly DetectiononVisA
Segmentation AUROC· 2025-03-04
98.9
best: 99.1 (Dinomaly ViT-L (model-unified multi-class))
Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection arXiv:2503.02424