List your company
ResearchFormerly WizWand

Out-of-Distribution Detection on Places with ImageNet-1k OOD In-distribution (test)

32.25FPR95Energy
29.58447.579565.57583.5705Oct 1, 2021Jun 22, 2022Mar 14, 2023Dec 4, 2023Aug 24, 2024May 16, 2025Feb 5, 2026
Updated 8mo ago

Evaluation results

MethodLinks
EnergyBackbone=CLIP-L, Training Protocol=Fine-tuned2022.1132.2593.01
ASH-BBackbone=ResNet2023.0833.4592.31
GradOrthBackbone=ResNet2023.0833.6791.78
ReActBackbone=ResNet-502021.1133.8591.58
ReActBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1133.8591.58
ReActBackbone=ResNet2023.0833.8591.58
MCMBackbone=CLIP-L, Training Protocol=Zero-shot2022.1135.4292
EnergyBackbone=CLIP-B, Training Protocol=Fine-tuned2022.1136.6491.82
DICE + ReActBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1136.8690.67
DICE + ReActBackbone=ResNet2023.0836.8690.67
VRA-PBackbone=ResNet2023.0837.8591.27
RankfeatMethods Space=Feature, Backbone=Google BiT-S, Blocks=Block42023.1138.2688.34
RankfeatMethods Space=Feature, Backbone=Google BiT-S, Blocks=Block3+42023.1139.3490.93
KNNIn-distribution image requirement=Required2024.0939.6191.02
ASH-SBackbone=ResNet2023.0839.7890.98
VRA-DNBackbone=ResNet2023.0839.9490.75
GradOrthBackbone=MobileNet2023.0840.2789.12
EnergyIn-distribution image requirement=Not required2024.0940.2989.32
MaxLogitIn-distribution image requirement=Not required2024.0941.1589.6
GradNormBackbone=ResNet2023.0843.4880.64
HFTTIn-distribution image requirement=Not required2024.0943.5490.26
MCMBackbone=CLIP-B, Training Protocol=Zero-shot2022.1144.6989.77
NPOSIn-distribution image requirement=Required2024.0945.2789.44
DICEBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1146.4987.48
DICEBackbone=ResNet2023.0846.4987.48
DICE + ReActBackbone=MobileNet2023.0846.7888.02
MCMIn-distribution image requirement=Not required2024.0947.4888.68
ASH-B@90Methods Space=Feature, Backbone=Google BiT-S2023.1147.7389.61
GAIA-AMethods Space=Gradient, Backbone=Google BiT-S2023.1148.5588.94
MOSBackbone=BiT, Training Protocol=Fine-tuned2022.1149.5489.06
ExGradBackbone=ResNet2023.0850.6274.27
ReActBackbone=MobileNet2021.1151.5686.64
ReActBackbone=MobileNet2023.0851.5686.64
ASH-BBackbone=MobileNet2023.0851.887.56
MahalanobisBackbone=MobileNet2021.1152.0981.65
MahalanobisBackbone=MobileNet2023.0852.0983.63
DICEBackbone=MobileNet2023.0853.1185.81
ReActBackbone=ResNet-502026.0254.587.9
Fort et al.Backbone=ViT-L, Training Protocol=Fine-tuned2022.1155.1486.48
ODINBackbone=MobileNet2021.1157.3684.71
ODINBackbone=MobileNet2023.0857.3684.71
Fort et al.Backbone=ViT-B, Training Protocol=Fine-tuned2022.1157.9985.24
ASH-SBackbone=MobileNet2023.0858.8484.73
MSPBackbone=CLIP-L, Training Protocol=Fine-tuned2022.1159.8684.1
GradNormMethod Space=Gradient, Backbone=ResNetv2-101, Pre-trained on=ImageNet-1k, Input Resolution=480x4802021.1060.8684.82
GradNormMethods Space=Gradient, Backbone=Google BiT-S2023.1160.8684.82
ASH-PBackbone=ResNet2023.0861.7985.58
ReActMethods Space=Feature, Backbone=Google BiT-S2023.1162.6687.83
MSPIn-distribution image requirement=Required2024.0963.580.19
GODINBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1163.783.81
EnergyBackbone=ResNet-502021.1164.9282.86
EnergyBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1164.9282.86
Energy scoreBackbone=ResNet2023.0864.9282.86
ASH-PBackbone=MobileNet2023.0866.5983.47
EnergyBackbone=ResNet-502026.0266.683.8
MaxLogitBackbone=ResNet-502026.0266.783.8
FDBDBackbone=ResNet-502026.0267.883.7
ODINBackbone=ResNet-502021.1167.8981.78
ODINBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1167.8981.78
ODINBackbone=ResNet2023.0867.8981.78
MSPBackbone=CLIP-B, Training Protocol=Fine-tuned2022.1167.980.14
EnergyBackbone=MobileNet2021.1169.3781.19
Energy scoreBackbone=MobileNet2023.0869.3781.19
CaptionMethod Category=Textual OE, Level=Caption2023.1070.5481.51
GAIA-ZMethods Space=Gradient, Backbone=Google BiT-S2023.1171.0281.16
WordMethod Category=Textual OE, Level=Word2023.1071.0681.65
MSPBackbone=ResNet-502026.0272.980.4
EnergyMethod Space=Output, Backbone=ResNetv2-101, Pre-trained on=ImageNet-1k, Input Resolution=480x4802021.1073.0281.37
EnergyMethods Space=Output, Backbone=Google BiT-S2023.1173.0281.37
ZOCIn-distribution image requirement=Not required2024.0973.0683.39
MSPBackbone=ResNet-502021.1173.9979.76
MSPBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1173.9979.76
Softmax scoreBackbone=ResNet2023.0873.9979.76
D-KNNBackbone=ResNet-502026.0274.876.7
ReActMethod Category=Post-hoc2023.107580.72
EnergyMethod Category=Post-hoc2023.1076.1580.23
ODINMethod Space=Output, Backbone=ResNetv2-101, Pre-trained on=ImageNet-1k, Input Resolution=480x4802021.1076.2780.67
ODINMethods Space=Output, Backbone=Google BiT-S2023.1176.2780.67
MSPMethod Category=Post-hoc2023.1076.6977.81
KNNMethods Space=Feature, Backbone=Google BiT-S2023.1177.0974.87
KNNMethod Category=Post-hoc2023.1079.274.34
MSPBackbone=MobileNet2021.1179.2376.27
Softmax scoreBackbone=MobileNet2023.0879.2376.27
KNNBackbone=ResNet-502026.028073.6
Desc.Method Category=Textual OE, Level=Description2023.1081.2877.4
MSPMethod Space=Output, Backbone=ResNetv2-101, Pre-trained on=ImageNet-1k, Input Resolution=480x4802021.1081.4476.76
MSPMethods Space=Output, Backbone=Google BiT-S2023.1181.4476.76
ViMBackbone=ResNet-502026.0283.877.9
ODINMethod Category=Post-hoc2023.108867.17
NecoBackbone=ResNet-502026.0288.666.3
MahalanobisMethod Space=Feature, Backbone=ResNetv2-101, Pre-trained on=ImageNet-1k, Input Resolution=480x4802021.1089.7564.46
MahalanobisMethods Space=Feature, Backbone=Google BiT-S2023.1189.7564.46
WDiscOODBackbone=ResNet-502026.0296.853.7
MahalanobisBackbone=ResNet-502026.0297.551
MahalanobisBackbone=ResNet-502021.1198.441.79
MahalanobisBackbone=ResNet-50, Pre-trained=ImageNet-1k2021.1198.441.79
MahalanobisBackbone=ResNet2023.0898.441.79
MahalanobisIn-distribution image requirement=Required2024.0998.5465.96
MahalanobisMethod Category=Post-hoc2023.1098.944.65