Forte : Finding Outliers with Representation Typicality Estimation

Benchmark Model Rank Results
out-of-distribution-detection-on-imagenet-1k-vs-inaturalistForte#1AUROC: 99.67 ± 00.03FPR95: 0.64 ± 00.06
out-of-distribution-detection-on-imagenet-1k-vs-nincoForte#1AUROC: 98.34 ± 00.09FPR@95: 5.18 ± 00.51
out-of-distribution-detection-on-cifar-10-vs-svhnForte#2AUROC: 99.84 ± 00.05FPR95: 00.00 ± 00.00
out-of-distribution-detection-on-imagenet-1k-vs-texturesForte#2AUROC: 98.04 ± 00.10FPR95: 5.61 ± 00.25
out-of-distribution-detection-on-imagenet-1k-vs-openimage-oForte#3AUROC: 96.73 ± 00.11FPR95: 11.77 ± 00.57
out-of-distribution-detection-on-cifar-10-vs-cifar-100Forte#4AUROC: 97.63 ± 00.15AUPR: 09.69 ± 01.08