List your company
ResearchFormerly WizWand

OOD Detection on LSUN-Resize (OOD) with CIFAR-10 (ID) (test)

0FPR@95Hopfield Boosting
-3.995222.972449.9476.9076Aug 1, 2023Dec 12, 2023Apr 23, 2024Sep 3, 2024Jan 14, 2025May 27, 2025Oct 7, 2025
Updated 3mo ago

Evaluation results

MethodLinks
Hopfield BoostingBackbone=ResNet-182024.05099.98
DOSBackbone=ResNet-182024.05099.99
DOEBackbone=ResNet-182024.050100
DivOEBackbone=ResNet-182024.05099.89
DALBackbone=ResNet-182024.05099.92
POEMBackbone=ResNet-182024.05099.88
EBO-OEBackbone=ResNet-182024.05099.98
MSP-OEBackbone=ResNet-182024.05099.96
Hopfield BoostingMethod type=OE, Augmentations=Weak, Auxiliary outlier data=true, Backbone=ResNet-182024.05099.98
MedixTraining Data=Pin and Pwild2025.100.0199.98
WOODSTraining Data=Pin and Pwild2025.100.1199.38
MixOEBackbone=ResNet-182024.050.1699.89
OETraining Data=Pin and Pwild2025.100.5498.84
GradOrthBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.082.3398.71
ODINBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.083.0999.02
GradNormBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.083.3898.87
DICEBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.083.9198.3
Energy (w/ OE)Training Data=Pin and Pwild2025.104.8598.62
ASH-SBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.084.9698.92
ASHTraining Data=Pin only2025.104.9698.92
VRA-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.085.898.69
GradPCA-BatchBackbone=ResNetV2-50 (BiT-M)2025.056.4898.67
ASH-PBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.087.9798.33
ASH-BBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.088.1398.54
MahalanobisBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.089.1497.09
Energy scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.089.2898.12
KNNBackbone=ResNetV2-50 (BiT-M)2025.059.7798.11
KNN+Training Data=Pin only2025.1011.2297.98
ReActBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.0811.4697.87
CSITraining Data=Pin only2025.1012.1598.01
GAIA-ZBackbone=ResNet-34 (TIMM)2025.0513.0697.32
NCIBackbone=ResNetV2-50 (BiT-M)2025.0514.196.71
MahalanobisBackbone=ResNetV2-50 (BiT-M)2025.0517.3697.04
GradPCA-VecBackbone=ResNetV2-50 (BiT-M)2025.0517.9496.82
GAIA-ABackbone=ResNet-34 (TIMM)2025.0518.1696.9
GradPCA (block 3)Backbone=ResNet-34 (TIMM)2025.0519.6396.54
GradPCA (block 4)Backbone=ResNetV2-50 (BiT-M)2025.0520.0296.49
Revisited PCABackbone=ResNetV2-50 (BiT-M)2025.0523.0894.16
Proj. GradsBackbone=ResNetV2-50 (BiT-M)2025.0523.6194.57
ODINTraining Data=Pin only2025.1026.6294.57
PALMMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.0526.8595.68
GradPCABackbone=ResNet-34 (TIMM)2025.0527.0995
Revisited PCABackbone=ResNet-34 (TIMM)2025.0527.2694.15
KNNTraining Data=Pin only2025.1027.5794.71
EnergyTraining Data=Pin only2025.1027.5894.24
EnergyBackbone=ResNet-34 (TIMM)2025.0527.6892.94
NCIBackbone=ResNet-34 (TIMM)2025.0528.2195.25
ODINBackbone=ResNet-34 (TIMM)2025.0528.5292.84
Max logitsBackbone=ResNet-34 (TIMM)2025.0528.5592.84
DICETraining Data=Pin only2025.1028.9393.56
ReActBackbone=ResNet-34 (TIMM)2025.0529.3791.89
GradOrthBackbone=ResNet-34 (TIMM)2025.0529.7194.05
KNNBackbone=ResNet-34 (TIMM)2025.0529.9995.33
GENMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.0530.6994.02
DICEBackbone=ResNet-34 (TIMM)2025.0530.9690.62
ASHMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.0531.594.04
EBOMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.0531.6493.9
MaxLogitMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.0532.6292.84
GradPCA+DICEBackbone=ResNet-34 (TIMM)2025.0533.3392.87
Kernel PCA (CoRP)Backbone=ResNet-34 (TIMM)2025.0533.5494.81
ReActTraining Data=Pin only2025.1033.6393.58
NPOSMethod type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.0534.394.78
MSPBackbone=ResNetV2-50 (BiT-M)2025.0535.8587.58
GAIA-ABackbone=ResNetV2-50 (BiT-M)2025.0535.8692.51
GradPCABackbone=ResNetV2-50 (BiT-M)2025.0536.7692.97
SSD+Method type=Training, Augmentations=Strong, Auxiliary outlier data=false, Backbone=ResNet-182024.0538.1890.39
GradPCA+DICEBackbone=ResNetV2-50 (BiT-M)2025.0541.0792.2
Proj. GradsBackbone=ResNet-34 (TIMM)2025.0541.6686.85
Softmax scoreBackbone=DenseNet, In-Distribution Dataset=CIFAR-102023.0842.194.51
ODINBackbone=ResNetV2-50 (BiT-M)2025.0542.2983.16
Max logitsBackbone=ResNetV2-50 (BiT-M)2025.0542.3283.15
GradPCA-BatchBackbone=ResNet-34 (TIMM)2025.0542.3293.29
GradPCA-VecBackbone=ResNet-34 (TIMM)2025.0542.4793.32
MahalanobisTraining Data=Pin only2025.1042.6293.23
MahalanobisBackbone=ResNet-34 (TIMM)2025.0545.4992.2
MSPBackbone=ResNet-34 (TIMM)2025.0545.991.82
EnergyBackbone=ResNetV2-50 (BiT-M)2025.0549.8780.69
MSPTraining Data=Pin only2025.1052.1591.37
Kernel PCA (CoRP)Backbone=ResNetV2-50 (BiT-M)2025.0555.9989.45
MSPMethod type=Post-hoc, Augmentations=Weak, Auxiliary outlier data=false, Backbone=ResNet-182024.0557.5887.27
GradOrthBackbone=ResNetV2-50 (BiT-M)2025.0578.280.29
GAIA-ZBackbone=ResNetV2-50 (BiT-M)2025.0597.7248.22
ReActBackbone=ResNetV2-50 (BiT-M)2025.0599.5938.91
DICEBackbone=ResNetV2-50 (BiT-M)2025.0599.8844.5