Out-of-Distribution Detection on Texture
98.92AUROCViM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViMModel=BiT-S-R101x1, Source=feat+logit2022.03 | 98.92 | 4.69 | — | |
| ViMBackbone=BiT, Pre-trained=ImageNet2023.02 | 98.92 | 4.69 | — | |
| VRA++Backbone=BiT, Pre-trained=ImageNet2023.02 | 98.76 | 5.02 | — | |
| GradOrthBackbone=ResNet-502026.02 | 98.06 | 11.19 | — | |
| Catalyst(m) + ReActBackbone=MobileNet-v22026.02 | 97.76 | 8.69 | — | |
| ResidualModel=BiT-S-R101x1, Source=feat2022.03 | 97.66 | 11.16 | — | |
| ASHBackbone=ResNet-502026.02 | 97.6 | 11.95 | — | |
| GradOrthBackbone=MobileNet-v22026.02 | 97.52 | 12.69 | — | |
| Catalyst(μ) + ReActBackbone=ResNet-502026.02 | 97.38 | 12.11 | — | |
| Catalyst(σ) + ReActBackbone=ResNet-502026.02 | 97.38 | 12.04 | — | |
| MahalanobisModel=BiT-S-R101x1, Source=feat+label2022.03 | 97.33 | 14.05 | — | |
| MahalanobisBackbone=BiT, Pre-trained=ImageNet2023.02 | 97.33 | 14.05 | — | |
| Catalyst(m) + ReActBackbone=ResNet-502026.02 | 97.31 | 12.06 | — | |
| Catalyst(σ) + ReActBackbone=MobileNet-v22026.02 | 97.31 | 10.18 | — | |
| KNN (α = 1%)Backbone=ResNet-502026.02 | 97.18 | 11.56 | — | |
| ASHBackbone=MobileNet-v22026.02 | 97.1 | 13.12 | — | |
| KNN (α = 100%)Backbone=ResNet-502026.02 | 97.07 | 11.77 | — | |
| ViMBackbone=ResNet-502026.02 | 96.83 | 14.88 | — | |
| Catalyst(μ) + ReActBackbone=MobileNet-v22026.02 | 96.83 | 13.6 | — | |
| SCALEBackbone=ResNet-502026.02 | 96.75 | 14.56 | — | |
| SCALEBackbone=MobileNet-v22026.02 | 96.65 | 14.79 | — | |
| Catalyst(σ)Backbone=MobileNet-v22026.02 | 96.37 | 14.02 | — | |
| Catalyst(m)Backbone=MobileNet-v22026.02 | 96.35 | 14.18 | — | |
| ReAct+DICEBackbone=MobileNet-v22026.02 | 96.33 | 16.03 | — | |
| Catalyst(σ)Backbone=ResNet-502026.02 | 95.94 | 15.85 | — | |
| Catalyst(m)Backbone=ResNet-502026.02 | 95.88 | 16.08 | — | |
| ViMModel=Res50d, Source=feat+logit2022.03 | 95.84 | 20.58 | — | |
| ViMModel=ViT-B/16, Source=feat+logit2022.03 | 95.34 | 20.31 | — | |
| ResidualModel=Res50d, Source=feat2022.03 | 94.62 | 25.89 | — | |
| MahalanobisModel=ViT-B/16, Source=feat+label2022.03 | 94.24 | 25.17 | — | |
| Catalyst(μ)Backbone=MobileNet-v22026.02 | 94.17 | 23.42 | — | |
| MahalanobisModel=Res50d, Source=feat+label2022.03 | 94.15 | 28.22 | — | |
| Catalyst(μ)Backbone=ResNet-502026.02 | 94.01 | 22.29 | — | |
| ViMModel=RepVGG, Source=feat+logit2022.03 | 93.69 | 23.76 | — | |
| Adaptive Multi-prompt Contrastive NetworkShot=82025.06 | 93.43 | 38.31 | — | |
| EnergyModel=ViT-B/16, Source=logit2022.03 | 93.39 | 28.22 | — | |
| ReActModel=ViT-B/16, Source=feat+logit2022.03 | 93.34 | 28.49 | — | |
| ReAct+DICEBackbone=ResNet-502026.02 | 93.33 | 28.4 | — | |
| ResidualModel=RepVGG, Source=feat2022.03 | 93.05 | 28.66 | — | |
| ODINModel=ViT-B/16, Source=prob+grad2022.03 | 93.01 | 30.6 | — | |
| MaxLogitModel=ViT-B/16, Source=logit2022.03 | 93.01 | 30.6 | — | |
| MahalanobisModel=RepVGG, Source=feat+label2022.03 | 92.71 | 32.03 | — | |
| ViMModel=Swin, Source=feat+logit2022.03 | 92.34 | 38.49 | — | |
| BATSBackbone=ResNet-502026.02 | 92.27 | 38.9 | — | |
| ResidualModel=ViT-B/16, Source=feat2022.03 | 92.21 | 33.8 | — | |
| fDBDBackbone=ResNet-502026.02 | 92.12 | 37.5 | — | |
| SCTShot=82025.06 | 91.82 | 40.35 | — | |
| LAPSBackbone=ResNet-502026.02 | 91.81 | 41.49 | — | |
| NN-GuideBackbone=ResNet-502026.02 | 91.52 | 24.93 | — | |
| DICEBackbone=MobileNet-v22026.02 | 91.46 | 32.57 | — | |
| PEFT‡Pre-trained model=OpenCLIP2025.09 | 91.32 | 38.26 | 94.66 | |
| ResidualModel=Swin, Source=feat2022.03 | 91.31 | 43.97 | — | |
| ReActModel=Res50d, Source=feat2022.03 | 91.12 | 39.26 | — | |
| GradNormBackbone=MobileNet-v22026.02 | 90.99 | 34.95 | — | |
| LoCoOpShot=82025.06 | 90.98 | 42.49 | — | |
| ReActBackbone=MobileNet-v22026.02 | 90.96 | 40.25 | — | |
| BATSBackbone=MobileNet-v22026.02 | 90.76 | 38.69 | — | |
| ReActModel=BiT-S-R101x1, Source=feat+logit2022.03 | 90.64 | 50.25 | — | |
| ReActBackbone=BiT, Pre-trained=ImageNet2023.02 | 90.64 | 50.25 | — | |
| DICEBackbone=ResNet-502026.02 | 90.48 | 32.38 | — | |
| ReActBackbone=ResNet-502026.02 | 90.3 | 46.33 | — | |
| MahalanobisModel=Swin, Source=feat+label2022.03 | 89.92 | 49.17 | — | |
| CoOpShot=82025.06 | 89.92 | 43.29 | — | |
| Clipped HNNTraining Dataset=CIFAR10, Scoring Function=softmax2021.07 | 89.91 | 47.12 | 97.39 | |
| Adaptive Multi-prompt Contrastive NetworkShot=12025.06 | 89.88 | 39.16 | — | |
| ViMBackbone=MobileNet-v22026.02 | 89.59 | 40.71 | — | |
| NN-GuideBackbone=MobileNet-v22026.02 | 89.32 | 38.78 | — | |
| LoCoOpShot=12025.06 | 89.13 | 49.25 | — | |
| NPOSShot=Full2025.06 | 88.8 | 46.12 | — | |
| KL MatchingModel=ViT-B/16, Source=prob2022.03 | 88.76 | 44.09 | — | |
| LAPSBackbone=MobileNet-v22026.02 | 88.29 | 51.37 | — | |
| PEFT†Pre-trained model=CLIP2025.09 | 87.86 | 49.45 | 92.79 | |
| ODINShot=Full2025.06 | 87.85 | 51.67 | — | |
| CoOpShot=12025.06 | 87.83 | 50.64 | — | |
| GradNormBackbone=ResNet-502026.02 | 87.73 | 38.15 | — | |
| ViMShot=Full2025.06 | 87.18 | 53.94 | — | |
| MSPModel=ViT-B/16, Source=prob2022.03 | 87.1 | 48.55 | — | |
| KL MatchingModel=BiT-S-R101x1, Source=prob2022.03 | 86.92 | 51.05 | — | |
| KL MatchingModel=Swin, Source=prob2022.03 | 86.89 | 52.93 | — | |
| EnergyBackbone=ResNet-502026.02 | 86.73 | 52.29 | — | |
| SCTShot=12025.06 | 86.66 | 48.87 | — | |
| EnergyBackbone=MobileNet-v22026.02 | 86.58 | 54.54 | — | |
| SeTARShot=02025.06 | 86.58 | 55.83 | — | |
| MCMShot=02025.06 | 86.11 | 57.77 | — | |
| KL MatchingModel=Res50d, Source=prob2022.03 | 86.07 | 61.36 | — | |
| ODINModel=Swin, Source=prob+grad2022.03 | 85.74 | 44.59 | — | |
| KNNShot=Full2025.06 | 85.67 | 64.35 | — | |
| ODINBackbone=ResNet-502026.02 | 85.62 | 50.23 | — | |
| ReActModel=Swin, Source=feat2022.03 | 85.51 | 47.91 | — | |
| MSPModel=Swin, Source=prob2022.03 | 85.21 | 51.9 | — | |
| ODINBackbone=MobileNet-v22026.02 | 85.03 | 49.96 | — | |
| MahalanobisBackbone=ResNet-502026.02 | 85.01 | 55.8 | — | |
| KL MatchingModel=DeiT, Source=prob2022.03 | 84.89 | 63.47 | — | |
| MaxLogitModel=Swin, Source=logit2022.03 | 84.67 | 47.42 | — | |
| ViMModel=DeiT, Source=feat+logit2022.03 | 84.42 | 73.02 | — | |
| GL-MCMShot=02025.06 | 83.63 | 57.93 | — | |
| MahalanobisModel=DeiT, Source=feat+label2022.03 | 83.58 | 77.31 | — | |
| KL MatchingModel=RepVGG, Source=prob2022.03 | 83.18 | 62.09 | — | |
| ENNTraining Dataset=CIFAR10, Scoring Function=softmax2021.07 | 82.8 | 65.04 | 94.59 | |
| MSPModel=Res50d, Source=prob2022.03 | 82.75 | 64.4 | — |