List your company
ResearchFormerly WizWand

Image Classification on ImageNet-C (mCE)

28.2mCEDINOv2
26.7236.7146.756.69Nov 11, 2019Dec 7, 2020Jan 3, 2022Jan 30, 2023Feb 26, 2024Mar 24, 2025Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
DINOv2Architecture=ViT-g/14, Pretraining Data=LVD-142M, Resolution=224, Protocol=linear probe on frozen features2023.0428.2
Noisy Student Trainingunlabeled data=300M images, backbone=EfficientNet2019.1128.3
CAFormer-B36Params (M)=99, MACS (G)=72.2, Resolution=384, Pre-training=ImageNet-21K2022.1030.8
MAE-H + DAT2022.0931.4
DINOv2Architecture=ViT-L/14, Pretraining Data=LVD-142M, Resolution=224, Protocol=linear probe on frozen features2023.0431.5
QUESTModel=ViT-L/16†, Epochs=400 + 20, Training Recipe=DeiT-32026.0332.3
StandardModel=ViT-L/16†, Epochs=400 + 20, Training Recipe=DeiT-32026.0332.5
QUESTModel=ViT-H/14†, Epochs=400 + 20, Training Recipe=DeiT-32026.0333.7
MAE-H2022.0933.92
StandardModel=ViT-H/14†, Epochs=400 + 20, Training Recipe=DeiT-32026.0334.1
QUESTModel=ViT-B/16, Epochs=400 + 20, Training Recipe=DeiT-32026.0335.7
StandardModel=ViT-B/16, Epochs=400 + 20, Training Recipe=DeiT-32026.0336.5
VOLO-D5HAT=true, Resolution=224x2242022.0438.4
Dr. ViTPre-trained=ImageNet-21K, Fine-tuned on=ImageNet, Resolution=384x384, Discrete representation=true2021.1138.74
Dr. ViTPre-trained=ImageNet-21K, Fine-tuned on=ImageNet, Resolution=512x512, Discrete representation=true2021.1138.97
EVT-XLParams(M)=205, FLOPs(G)=36.42026.0441.1
VOLO-D5HAT=false, Resolution=224x2242022.0441.8
ViT-BHAT=true, Resolution=224x2242022.0442.2
DINOv2Architecture=ViT-B/14, Pretraining Data=LVD-142M, Resolution=224, Protocol=linear probe on frozen features2023.0442.7
EVT-LParams(M)=101, FLOPs(G)=18.22026.0442.7
QUESTModel=ViT-B/16, Epochs=100, Training Recipe=DeiT-12026.0342.9
GC ViT-LParams(M)=201, FLOPs(G)=32.62026.0442.9
EVT-BParams(M)=57, FLOPs(G)=9.82026.0443.1
RVT-B-AdaNCAParams (M)=91, FLOPS (G)=19, Pre-training=ImageNet-1K2024.0643.2
QUESTModel=ViT-S/16, Epochs=200, Training Recipe=DeiT-12026.0343.2
TransNeXt-BaseParams(M)=90, FLOPs(G)=18.42026.0443.5
DATBackbone=ViT-B/16, Pre-training=ImageNet-21K, Fine-tuning=ImageNet-1K, Input Resolution=224x2242023.1143.6
VOLO-D1HAT=true, Resolution=224x2242022.0443.7
TAPADL-FANParams (M)=50.7, FLOPS (G)=11.8, Pre-training=ImageNet-1K2024.0643.7
iBOTArchitecture=ViT-L/16, Pretraining Data=INet-22k, Resolution=224, Protocol=linear probe on frozen features2023.0443.9
TransNeXt-SmallParams(M)=50, FLOPs(G)=10.32026.0443.9
EVT-SParams(M)=27, FLOPs(G)=4.62026.0444.2
ConViT-B-AdaNCAParams (M)=89, FLOPS (G)=19, Pre-training=ImageNet-1K2024.0644.3
QKNorm-DS‡Model=ViT-B/16, Epochs=100, Training Recipe=DeiT-12026.0344.4
AugReg-ViT + DAT2022.0944.65
TAPADL-RVTParams (M)=89.4, FLOPS (G)=17.9, Pre-training=ImageNet-1K2024.0644.7
FAN-B-AdaNCAParams (M)=51.7, FLOPS (G)=12.4, Pre-training=ImageNet-1K2024.0644.7
StandardModel=ViT-S/16, Epochs=200, Training Recipe=DeiT-12026.0344.8
ConViT-BParams (M)=93.6, FLOPS (G)=19.2, Pre-training=ImageNet-22K2024.0645.2
OpenCLIPArchitecture=ViT-G/14, Pretraining Data=LAION-2B, Resolution=224, Protocol=linear probe on frozen features2023.0445.3
Prev. SOTAweakly labeled data=3.5B Instagram images2019.1145.7
Dr. ViTPre-trained=ImageNet-21K, Fine-tuned on=ImageNet, Resolution=384x384, Discrete representation=discrete only2021.1145.86
FAN-BParams (M)=50.4, FLOPS (G)=11.7, Pre-training=ImageNet-1K2024.0646.1
CAFormer-S18Params (M)=26, MACS (G)=13.4, Resolution=384, Pre-training=ImageNet-1K2022.1046.1
Dr. ViTBackbone=ViT-B, Discrete Only=false2021.1146.22
DrViT2022.0946.22
Dr. ViTBackbone=ViT-B, Resolution=384x3842021.1146.32
ViT-BHAT=false, Resolution=224x2242022.0446.4
MaxViT-SmallParams(M)=69, FLOPs(G)=11.72026.0446.4
TransNeXt-TinyParams(M)=28, FLOPs(G)=5.72026.0446.5
VOLO-D1HAT=false, Resolution=224x2242022.0446.8
RVT-BParams (M)=88.5, FLOPS (G)=17.7, Pre-training=ImageNet-1K2024.0646.8
ConvNeXt-BParams(M)=89, FLOPs(G)=15.42026.0446.8
Swin-BHAT=true, Resolution=224x2242022.0446.9
ConViT-BParams (M)=86.5, FLOPS (G)=17.7, Pre-training=ImageNet-1K2024.0646.9
BiFormer-BParams(M)=57, FLOPs(G)=9.82026.0447.2
BiFormer-SParams(M)=26, FLOPs(G)=4.52026.0448.5
Swin-SHAT=true, Resolution=224x2242022.0448.6
EVT-TParams(M)=15, FLOPs(G)=2.52026.0449.2
ViT-BPre-trained=ImageNet-21K, Fine-tuned on=ImageNet, Resolution=384x384, Discrete representation=false2021.1149.62
ViT-SHAT=true, Resolution=224x2242022.0449.7
QUESTModel=ViT-L/16, Epochs=100, Training Recipe=DeiT-12026.0350.3
TransNeXt-MicroParams(M)=13, FLOPs(G)=2.72026.0450.8
DeepAugment + Augmix + DAT2022.0950.82
Swin-B-AdaNCAParams (M)=90.7, FLOPS (G)=16.3, Pre-training=ImageNet-1K2024.0651.5
Swin-BHAT=false, Resolution=224x2242022.0451.7
ConvFormer-S18Params (M)=27, MACS (G)=3.9, Resolution=224, Pre-training=ImageNet-1K2022.1051.7
Swin-SHAT=false, Resolution=224x2242022.0451.8
DADBackbone=ViT-B/16, Pre-training=ImageNet-21K, Fine-tuning=ImageNet-1K, Input Resolution=224x2242023.1152
Swin-BParams (M)=94.1, FLOPS (G)=16.7, Pre-training=ImageNet-22K2024.0653.2
ViT-SHAT=false, Resolution=224x2242022.0453.3
ViT-BBackbone=ViT-B2021.1153.51
DeepAugment + Augmix2022.0953.55
DeepAugment + AugMixBackbone=ResNet-502020.1053.6
Swin-THAT=true, Resolution=224x2242022.0453.9
Swin-BParams (M)=87.8, FLOPS (G)=15.4, Pre-training=ImageNet-1K2024.0654.3
DINOv2Architecture=ViT-S/14, Pretraining Data=LVD-142M, Resolution=224, Protocol=linear probe on frozen features2023.0454.4
QKNorm-DS‡Model=ViT-L/16, Epochs=100, Training Recipe=DeiT-12026.0354.4
AugReg-ViT2022.0954.5
Swin-BParams(M)=88, FLOPs(G)=15.42026.0454.5
FocalNet-TParams(M)=29, FLOPs(G)=4.52026.0455
Dr. ViTBackbone=ViT-B, Discrete Only=true2021.1155.21
SWAD2021.0255.7
BiFormer-TParams(M)=13, FLOPs(G)=2.22026.0455.7
DINOArchitecture=ViT-B/8, Pretraining Data=INet-1k, Resolution=224, Protocol=linear probe on frozen features2023.0456.6
SWA2021.0256.8
Dr. ViTBackbone=ViT-S, Discrete Only=false2021.1156.89
AENIBBackbone=ViT-B/162023.0357.5
ERM2021.0257.6
Swin-THAT=false, Resolution=224x2242022.0458
BaselineBackbone=ViT-B/162023.0358.6
ViT-BBackbone=ViT-B, Resolution=384x3842021.1159.01
DeepAugment2022.0960.37
MAEArchitecture=ViT-H/14, Pretraining Data=INet-1k, Resolution=224, Protocol=linear probe on frozen features2023.0461.4
ViTBackbone=ViT-B/16, Pre-training=ImageNet-21K, Fine-tuning=ImageNet-1K, Input Resolution=224x2242023.1161.9
ViT-SBackbone=ViT-S2021.1161.99
Swin-TParams (M)=29, MACS (G)=4.5, Resolution=224, Pre-training=ImageNet-1K2022.1062
Swin-TParams(M)=29, FLOPs(G)=4.52026.0462
PVTv2-B1Params(M)=14, FLOPs(G)=2.12026.0462.6
AENIBBackbone=ViT-S/162023.0365.2