List your company
ResearchFormerly WizWand

OOD Detection on iNaturalist OOD

99.16AUROCGradPCA
50.789663.347375.90588.4627Sep 20, 2022Mar 1, 2023Aug 10, 2023Jan 20, 2024Jun 30, 2024Dec 9, 2024May 21, 2025
Updated 8mo ago

Evaluation results

MethodLinks
GradPCABackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0599.163.74
MahalanobisBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0599.094.75
EnergyBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0597.911.29
ASH-SBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0997.8711.49
KNNBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0597.6312.48
Max logitsBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0597.5413.27
ODINBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0597.5413.27
ASH-BBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0997.3214.21
ReActBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0597.2914.9
DICEBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0596.7315.2
DICE + ReActBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0996.2418.64
ReActBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0996.2220.38
CADetTraining=Supervised, Backbone=ResNet502022.1095.28—
MSPBackbone=ViT-B/16, In-Distribution Dataset=ImageNet-1k2025.0594.9222.05
DICEBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0994.4925.63
ASH-BBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0994.2831.46
DICE + ReActBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0993.5732.3
ASH-PBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0992.5144.57
ASH-SBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0991.9439.1
ReActBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0991.5342.4
DICEBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0990.8343.09
KL MatchingTraining=Supervised, Backbone=ResNet502022.1090.48—
ASH-PBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0990.4654.92
Energy scoreBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0989.9555.72
ODINBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0989.6647.66
MahalanobisTraining=Supervised, Backbone=ResNet502022.1089.48—
ViMTraining=Supervised, Backbone=ResNet502022.1089.26—
Energy scoreBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0988.9159.5
MSPTraining=Supervised, Backbone=ResNet502022.1088.58—
Softmax scoreBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0987.7454.99
ODINBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0987.6255.39
ReActTraining=Supervised, Backbone=ResNet502022.1087.27—
ODINTraining=Supervised, Backbone=ResNet502022.1086.48—
MaxLogitTraining=Supervised, Backbone=ResNet502022.1086.42—
Softmax scoreBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.0985.3264.29
ResidualTraining=Supervised, Backbone=ResNet502022.1084.63—
CADetTraining=Self-supervised (contrastive), Backbone=ResNet502022.1083.42—
MahalanobisBackbone=MobileNet, In-Distribution Dataset=ImageNet-1k2022.098162.11
EnergyTraining=Supervised, Backbone=ResNet502022.1080.5—
MahalanobisBackbone=ResNet, In-Distribution Dataset=ImageNet-1k2022.0952.6597