Out-of-Distribution Detection on ImageNet-1k Textures ID OOD
97.97AUROCVIM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VIMBackbone=ResNet-502023.06 | 97.97 | — | |
| NAC-UEBackbone=ResNet-502023.06 | 97.9 | — | |
| KNNBackbone=ResNet-502026.02 | 97.2 | 11.9 | |
| KNNBackbone=ResNet-502023.06 | 97.09 | — | |
| MM++Backbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc, K=22026.06 | 97 | 14.52 | |
| ASHBackbone=ResNet-502023.06 | 96.9 | — | |
| ViMBackbone=ResNet-502026.02 | 96.8 | 15.2 | |
| X-MahalanobisBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Fine-tuning2026.06 | 96.65 | 11.7 | |
| InterNegArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 96.26 | 21.85 | |
| NecoBackbone=ResNet-502026.02 | 95.7 | 17.5 | |
| D-KNNBackbone=ResNet-502026.02 | 95.6 | 15.9 | |
| ASH-B@90Model=DenseNet-1212022.09 | 95.38 | 22.75 | |
| ASH-S@90Model=DenseNet-1212022.09 | 95.01 | 22.04 | |
| AdaNegBackbone=ViT-B/16 CLIP, Training Status=Zero-Shot Training-free2024.10 | 94.93 | 31.27 | |
| AdaNegArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 94.93 | 31.27 | |
| SUPREMEArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 94.45 | 26.77 | |
| ID-LikeArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 94.32 | 25.27 | |
| NAC-UEBackbone=ViT-b162023.06 | 94.17 | — | |
| CSPArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 93.86 | 25.52 | |
| SHEBackbone=ResNet-502023.06 | 93.6 | — | |
| ReActBackbone=ResNet-502023.06 | 92.79 | — | |
| SHEBackbone=ViT-b162023.06 | 92.65 | — | |
| ASH-B@65Model=VGG-162022.09 | 92.35 | 30.25 | |
| Local-PromptArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 92.29 | 34.5 | |
| FDBDBackbone=ResNet-502026.02 | 91.9 | 38.1 | |
| NegPromptBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 91.6 | 35.21 | |
| WDiscOODBackbone=ResNet-502026.02 | 91.6 | 38.8 | |
| NegPromptArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 91.6 | 35.21 | |
| CLIPScopeArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 91.41 | 38.37 | |
| KNNBackbone=ViT-b162023.06 | 91.12 | — | |
| LAPTBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 91.06 | 40.32 | |
| LAPTArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 91.06 | 40.32 | |
| ASH-S@95Model=VGG-162022.09 | 90.95 | 39.08 | |
| CLIPNBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 90.93 | 40.83 | |
| CLIPNArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 90.93 | 40.83 | |
| VIMBackbone=ViT-b162023.06 | 90.61 | — | |
| LSNBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 90.42 | 38.54 | |
| LSNArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 90.42 | 38.54 | |
| GENBackbone=ViT-b162023.06 | 90.23 | — | |
| NegLabelBackbone=ViT-B/16 CLIP, Training Status=Zero-Shot Training-free2024.10 | 90.22 | 43.56 | |
| NegLabelArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 90.22 | 43.56 | |
| LoCoOpBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 90.19 | 42.28 | |
| LoCoOpArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 90.19 | 42.28 | |
| CoVerArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 90.14 | 43.39 | |
| ReActBackbone=ResNet-502026.02 | 90 | 46.6 | |
| Energy scoreModel=VGG-162022.09 | 89.98 | 44.24 | |
| ASH-S@90Model=VGG-162022.09 | 89.96 | 44.43 | |
| MDSBackbone=ResNet-502023.06 | 89.8 | — | |
| MahalanobisBackbone=ResNet-502026.02 | 89.7 | 44.8 | |
| KNNBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 89.56 | 42.93 | |
| MDSBackbone=ViT-b162023.06 | 89.41 | — | |
| RMDSBackbone=ViT-b162023.06 | 89.38 | — | |
| Relative Mahalanobis++Backbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 89.31 | 48.26 | |
| Mahalanobis++Backbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 89.26 | 48.26 | |
| NPOSBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 88.8 | 46.12 | |
| OpenMaxBackbone=ResNet-502023.06 | 88.1 | — | |
| ODINBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 87.85 | 51.67 | |
| ODINArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 87.85 | 51.67 | |
| GENBackbone=ResNet-502023.06 | 87.59 | — | |
| ViMBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 87.18 | 53.94 | |
| ViMArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 87.18 | 53.94 | |
| Relative MahalanobisBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 86.89 | 58.79 | |
| EnergyBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 86.76 | 57.61 | |
| EnergyArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 86.76 | 57.61 | |
| ReActBackbone=ViT-b162023.06 | 86.66 | — | |
| EnergyBackbone=ResNet-502026.02 | 86.6 | 53.1 | |
| ASH-S@99Model=VGG-162022.09 | 86.36 | 43.6 | |
| VOSBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 86.33 | 61.02 | |
| VOSArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 86.33 | 61.02 | |
| MaxLogitBackbone=ResNet-502026.02 | 86.2 | 55.6 | |
| MCMBackbone=ViT-B/16 CLIP, Training Status=Zero-Shot Training-free2024.10 | 86.12 | 58.5 | |
| MCMArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 86.12 | 58.5 | |
| RMDSBackbone=ResNet-502023.06 | 86.08 | — | |
| KNNBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 85.67 | 64.35 | |
| KNNArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 85.67 | 64.35 | |
| EOEArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 85.64 | 57.53 | |
| ReActBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 85.53 | 44.45 | |
| OpenMaxBackbone=ViT-b162023.06 | 85.52 | — | |
| MahalanobisBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 85.43 | 64.57 | |
| Energy scoreModel=DenseNet-1212022.09 | 85.42 | 52.11 | |
| EnergyBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 83.53 | 47.07 | |
| ASH-B@65Model=DenseNet-1212022.09 | 83.1 | 56.21 | |
| MSPBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 82.72 | 58.39 | |
| MSPBackbone=ResNet-502026.02 | 80.4 | 67.1 | |
| MSPBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 79.69 | 71.93 | |
| MSPArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 79.69 | 71.93 | |
| ZOCBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 76.46 | 90.9 | |
| ZOCArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=requiring training on ID or extra data2026.03 | 76.46 | 98.9 | |
| ODINBackbone=ViT-B/16, Pre-training=ImageNet-21K, Evaluation Protocol=Post-hoc2026.06 | 72.78 | 65.5 | |
| GradNormBackbone=ViT-B/16 CLIP, Training Status=Methods requiring training2024.10 | 70.26 | 79.36 | |
| GradNormArchitecture=CLIP ViT-B/16, Modality=Visual-based, Training Requirement=requiring training on ID or extra data2026.03 | 70.26 | 79.36 | |
| EnergyBackbone=ViT-B/16 CLIP, Training Status=Zero-Shot Training-free2024.10 | 65.56 | 93.65 | |
| EnergyArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 65.56 | 93.65 | |
| MahalanobisBackbone=ViT-B/16 CLIP, Training Status=Zero-Shot Training-free2024.10 | 64.23 | 98.46 | |
| MahalanobisArchitecture=CLIP ViT-B/16, Modality=VLM-based, Training Requirement=no training on ID or extra data2026.03 | 64.23 | 98.46 | |
| ASHBackbone=ViT-b162023.06 | 48.53 | — |