Reverse Distillation

Reported on 6 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.

Methodology6 results

Anomaly DetectiononFashion-MNIST
ROC AUC· 2022-01-26
95
best: 98.6 (GAN based Anomaly Detection in Imbalance Problems)
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703
Anomaly DetectiononOne-class CIFAR-10
AUROC· 2022-01-26
86.5
best: 99.6 (CLIP (OE))
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703
Anomaly DetectiononMVTec AD
Detection AUROC· 2022-01-26
98.5
best: 99.9 (GLASS)
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703
Anomaly DetectiononMVTec AD
Segmentation AUPRO· 2022-01-26
93.9
best: 98.4 (WeakREST-Block)
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703
Anomaly DetectiononMVTec AD
Segmentation AUROC· 2022-01-26
97.8
best: 99.7 (WeakREST-Block)
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703
Anomaly DetectiononVisA
Segmentation AUPRO (until 30% FPR)· 2022-01-26
70.9
best: 96.1 (AnomalyDINO-S (full-shot))
Anomaly Detection via Reverse Distillation from One-Class Embedding arXiv:2201.10703