Image Classification on DTD
88.9AccuracyiSICE
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| iSICEBackbone=ConvNext-T2023.04 | 88.9 | — | — | — | — | — | — | — | |
| iSICEBackbone=Swin-T2023.04 | 88.3 | — | — | — | — | — | — | — | |
| Precision ΩBackbone=Swin-T2023.04 | 86.5 | — | — | — | — | — | — | — | |
| DeepCOVBackbone=VGG-162023.04 | 86.3 | — | — | — | — | — | — | — | |
| DeepKSPDBackbone=VGG-162023.04 | 86.3 | — | — | — | — | — | — | — | |
| iSQRT-COVBackbone=Swin-T2023.04 | 86.1 | — | — | — | — | — | — | — | |
| Florence-CoSwin-HResolution=384px, Backbone=CoSwin-H, Evaluation Protocol=Linear Probing2021.11 | 86 | — | — | — | — | — | — | — | |
| iSQRT-COVBackbone=ConvNext-T2023.04 | 84.7 | — | — | — | — | — | — | — | |
| DeepCOV-ResNetBackbone=ResNet-502023.04 | 84.6 | — | — | — | — | — | — | — | |
| Auto-FlexSwitch+MSE (λ = 0.05)Type=Dynamic, Storage (MB)=98.41, Backbone=ViT-L/142026.04 | 84.57 | — | — | — | — | — | — | — | |
| Auto-FlexSwitch+KL (λ = 0.3)Type=Dynamic, Storage (MB)=80.69, Backbone=ViT-L/142026.04 | 84.47 | — | — | — | — | — | — | — | |
| Auto-FlexSwitch+CKA (λ = 3)Type=Dynamic, Storage (MB)=89.76, Backbone=ViT-L/142026.04 | 84.26 | — | — | — | — | — | — | — | |
| IndividualBackbone=ViT-L/142026.04 | 84.15 | — | — | — | — | — | — | — | |
| BCNNBackbone=VGG-162023.04 | 84 | — | — | — | — | — | — | — | |
| CBPBackbone=VGG-162023.04 | 84 | — | — | — | — | — | — | — | |
| Precision ΩBackbone=ConvNext-T2023.04 | 83.9 | — | — | — | — | — | — | — | |
| Multiple modelsBackbone=ViT-L/142022.08 | 83.7 | — | — | — | — | — | — | — | |
| Joint patchingBackbone=ViT-L/142022.08 | 83.6 | — | — | — | — | — | — | — | |
| Auto-SwitchType=Dynamic, Storage (MB)=556.88, Backbone=ViT-L/142026.04 | 83.51 | — | — | — | — | — | — | — | |
| CLIP-ViT-L/14Resolution=336px, Evaluation Protocol=Linear Probing2021.11 | 83 | — | — | — | — | — | — | — | |
| EMR-MergingType=Dynamic, Storage (MB)=1633.46, Backbone=ViT-L/142026.04 | 82.71 | — | — | — | — | — | — | — | |
| EfficientNet-L2Resolution=800px, Evaluation Protocol=Linear Probing2021.11 | 82.5 | — | — | — | — | — | — | — | |
| Multiple modelsBackbone=ViT-B/162022.08 | 82.3 | — | — | — | — | — | — | — | |
| Twin-MergingType=Dynamic, Storage (MB)=1780.62, Backbone=ViT-L/142026.04 | 82.29 | — | — | — | — | — | — | — | |
| DINOv2-LVD142MBackbone=ViT-L/14, Training Scale=large scale datasets, Evaluation protocol=Linear probe2026.06 | 82.23 | — | — | — | — | — | — | — | |
| CLIP-ResNet-50x64Evaluation Protocol=Linear Probing2021.11 | 82 | — | — | — | — | — | — | — | |
| CERSABackbone=ViT-Base, α=0.95, β=0.952026.05 | 81.2 | — | — | — | — | — | — | — | |
| VISReg-Inet22KBackbone=ViT-L/14, Training Scale=large scale datasets, Evaluation protocol=Linear probe2026.06 | 80.74 | — | — | — | — | — | — | — | |
| SimCLRv2Backbone=ResNet-152x3, Evaluation Protocol=Linear Probing2021.11 | 80.5 | — | — | — | — | — | — | — | |
| SVFitBackbone=ViT-Base, Rank=7682026.05 | 80.5 | — | — | — | — | — | — | — | |
| Precision ΩBackbone=Swin-B2023.04 | 80.1 | — | — | — | — | — | — | — | |
| X-FM_baseLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.01 | 80 | — | — | — | — | — | — | — | |
| iSICEBackbone=Swin-B2023.04 | 79.8 | — | — | — | — | — | — | — | |
| SVFTBackbone=ViT-Base2026.05 | 79.8 | — | — | — | — | — | — | — | |
| Joint patchingBackbone=ViT-B/162022.08 | 79.5 | — | — | — | — | — | — | — | |
| iSQRT-COVBackbone=Swin-B2023.04 | 79.4 | — | — | — | — | — | — | — | |
| IndividualBackbone=CLIP ViT-B/32, Merging Category=Non-merging2026.05 | 79.4 | — | — | — | — | — | — | — | |
| PyramidCLIPImage Encoder=ViT-B/16, Pretrain Dataset Size=143M, Evaluation Protocol=Linear probe2022.04 | 79.3 | — | — | — | — | — | — | — | |
| NNCLRArch=ViT-B/82021.04 | 79.2 | — | — | — | — | — | — | — | |
| CLIP*Image Encoder=ViT-B/16, Pretrain Dataset Size=400M, Evaluation Protocol=Linear probe2022.04 | 79.2 | — | — | — | — | — | — | — | |
| CLIPLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.01 | 79.2 | — | — | — | — | — | — | — | |
| DF-MergeType=Static, Backbone=ViT-L/142026.04 | 79.2 | — | — | — | — | — | — | — | |
| Traditional MTLBackbone=ViT-L/142026.04 | 79.1 | — | — | — | — | — | — | — | |
| AdaMerging++Type=Static, Backbone=ViT-L/142026.04 | 78.98 | — | — | — | — | — | — | — | |
| AdaMergingType=Static, Backbone=ViT-L/142026.04 | 78.94 | — | — | — | — | — | — | — | |
| SimCLRProtocol=Linear evaluation, Backbone=ResNet-50 (4x), Pre-training=ImageNet2020.02 | 78.9 | — | — | — | — | — | — | — | |
| SupervisedProtocol=Fine-tuned, Backbone=ResNet-50 (4x), Pre-training=ImageNet2020.02 | 78.8 | — | — | — | — | — | — | — | |
| TIESBackbone=ViT-L/16, Number of samples=5002026.06 | 78.8 | — | — | — | — | — | — | — | |
| MoW-MergingType=Dynamic, Storage (MB)=5159.51, Backbone=ViT-L/142026.04 | 78.78 | — | — | — | — | — | — | — | |
| SupervisedProtocol=Linear evaluation, Backbone=ResNet-50 (4x), Pre-training=ImageNet2020.02 | 78.7 | — | — | — | — | — | — | — | |
| Multiple modelsBackbone=ViT-B/322022.08 | 78.7 | — | — | — | — | — | — | — | |
| iSICEBackbone=ResNeXt-1012023.04 | 78.7 | — | — | — | — | — | — | — | |
| PiSSABackbone=ViT-Base, Rank=322026.05 | 78.7 | — | — | — | — | — | — | — | |
| TaDABackbone=ViT-L/16, Number of samples=5002026.06 | 78.7 | — | — | — | — | — | — | — | |
| DARE-TIESBackbone=ViT-L/16, Number of samples=5002026.06 | 78.6 | — | — | — | — | — | — | — | |
| Mag. PruningBackbone=ViT-L/16, Number of samples=5002026.06 | 78.4 | — | — | — | — | — | — | — | |
| DaVinciLinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.01 | 78.3 | — | — | — | — | — | — | — | |
| NNCLR + SupArch=ViT-B/82021.04 | 78.1 | — | — | — | — | — | — | — | |
| LinearBackbone=ViT-L/16, Number of samples=5002026.06 | 78 | — | — | — | — | — | — | — | |
| DARE-LinBackbone=ViT-L/16, Number of samples=5002026.06 | 78 | — | — | — | — | — | — | — | |
| Traditional MTLBackbone=CLIP ViT-B/32, Merging Category=Non-merging2026.05 | 77.9 | — | — | — | — | — | — | — | |
| SVD MergeBackbone=ViT-L/16, Number of samples=5002026.06 | 77.9 | — | — | — | — | — | — | — | |
| SimCLRProtocol=Fine-tuned, Backbone=ResNet-50 (4x), Pre-training=ImageNet2020.02 | 77.8 | — | — | — | — | — | — | — | |
| FTBackbone=ViT-Base2026.05 | 77.7 | — | — | — | — | — | — | — | |
| NNCLRArch=ViT-B/162021.04 | 77.3 | — | — | — | — | — | — | — | |
| PyramidCLIPImage Encoder=ViT-B/32, Pretrain Dataset Size=143M, Evaluation Protocol=Linear probe2022.04 | 77.3 | — | — | — | — | — | — | — | |
| FLAVALinear evaluation=true, Model size=Base, Patch size=16*16, Resolution=224*2242023.01 | 77.3 | — | — | — | — | — | — | — | |
| Task Arith.Backbone=ViT-L/16, Number of samples=5002026.06 | 77 | — | — | — | — | — | — | — | |
| Precision ΩBackbone=ResNeXt-1012023.04 | 76.9 | — | — | — | — | — | — | — | |
| CaCoBackbone=ResNet-50, Pre-training Dataset=ImageNet-1K, Pre-training Epochs=8002022.03 | 76.8 | — | — | — | — | — | — | — | |
| DeCLIPTask=Linear Probe, Pretrain Dataset=88M, Image Encoder=ResNet502022.04 | 76.8 | — | — | — | — | — | — | — | |
| Base modelBackbone=ViT-L/16, Number of samples=5002026.06 | 76.8 | — | — | — | — | — | — | — | |
| Task onlyBackbone=ViT-L/16, Number of samples=5002026.06 | 76.8 | — | — | — | — | — | — | — | |
| TWISTEvaluation Protocol=Linear evaluation, Backbone=ResNet-50, Pre-trained=ImageNet2021.10 | 76.6 | — | — | — | — | — | — | — | |
| VISRegBackbone=ViT-L/14, Heuristics=Excluded, Evaluation protocol=Linear probe2026.06 | 76.54 | — | — | — | — | — | — | — | |
| SimCLR v2Evaluation Protocol=Linear eval, Backbone=ResNet, Pre-training Dataset=ImageNet2020.11 | 76.5 | — | — | — | — | — | — | — | |
| CLIP*Image Encoder=ViT-B/32, Pretrain Dataset Size=400M, Evaluation Protocol=Linear probe2022.04 | 76.5 | — | — | — | — | — | — | — | |
| Seq. patchingBackbone=ViT-L/14, Seed=22022.08 | 76.5 | — | — | — | — | — | — | — | |
| TWISTEvaluation Protocol=Fine-tune, Backbone=ResNet-50, Pre-trained=ImageNet2021.10 | 76.4 | — | — | — | — | — | — | — | |
| CLIPTask=Linear Probe, Pretrain Dataset=400M, Image Encoder=ResNet502022.04 | 76.4 | — | — | — | — | — | — | — | |
| FNCEvaluation Protocol=Linear eval, Backbone=ResNet, Pre-training Dataset=ImageNet2020.11 | 76.3 | — | — | — | — | — | — | — | |
| BYOLArchitecture=ResNet-50, Pre-training Dataset=ImageNet, Evaluation Protocol=Fine-tuned2020.06 | 76.2 | — | — | — | — | — | — | — | |
| BYOLEvaluation Protocol=Fine-tune, Backbone=ResNet-50, Pre-trained=ImageNet2021.10 | 76.2 | — | — | — | — | — | — | — | |
| BYOLEvaluation Protocol=Fine-tuned, Backbone=ResNet, Pre-training Dataset=ImageNet2020.11 | 76.2 | — | — | — | — | — | — | — | |
| FNCEvaluation Protocol=Fine-tuned, Backbone=ResNet, Pre-training Dataset=ImageNet2020.11 | 76 | — | — | — | — | — | — | — | |
| Joint patchingBackbone=ViT-B/322022.08 | 76 | — | — | — | — | — | — | — | |
| PyramidCLIPTask=Linear Probe, Pretrain Dataset=143M, Image Encoder=ResNet502022.04 | 75.8 | — | — | — | — | — | — | — | |
| SimCLR (repro)Architecture=ResNet-50, Pre-training Dataset=ImageNet, Evaluation Protocol=Linear evaluation2020.06 | 75.7 | — | — | — | — | — | — | — | |
| SimCLRBackbone=ResNet-50, Pre-training=ImageNet2021.04 | 75.7 | — | — | — | — | — | — | — | |
| SimCLRArch=R502021.04 | 75.7 | — | — | — | — | — | — | — | |
| SimCLREpochs=1000, Backbone=ResNet-50, Pre-training Dataset=ImageNet2022.03 | 75.7 | — | — | — | — | — | — | — | |
| SimCLRBackbone=ResNet-50, Pre-training Dataset=ImageNet-1K, Pre-training Epochs=8002022.03 | 75.7 | — | — | — | — | — | — | — | |
| iSQRT-COVBackbone=ResNeXt-1012023.04 | 75.7 | — | — | — | — | — | — | — | |
| VISRegBackbone=ViT-B/16, Heuristics=Excluded, Evaluation protocol=Linear probe2026.06 | 75.69 | — | — | — | — | — | — | — | |
| Domain onlyBackbone=ViT-L/16, Number of samples=5002026.06 | 75.6 | — | — | — | — | — | — | — | |
| BYOLArchitecture=ResNet-50, Pre-training Dataset=ImageNet, Evaluation Protocol=Linear evaluation2020.06 | 75.5 | — | — | — | — | — | — | — | |
| BYOLBackbone=ResNet-50, Pre-training=ImageNet2021.04 | 75.5 | — | — | — | — | — | — | — | |
| NNCLRBackbone=ResNet-50, Pre-training=ImageNet2021.04 | 75.5 | — | — | — | — | — | — | — | |
| BYOLArch=R502021.04 | 75.5 | — | — | — | — | — | — | — | |
| NNCLRArch=R502021.04 | 75.5 | — | — | — | — | — | — | — |