List your company
ResearchFormerly WizWand

Image Classification on ImageNet-R 1.0 (val)

71.1Top-1 AccFAN-L-Hybrid
17.5431.44545.3559.255May 17, 2021Aug 7, 2021Oct 29, 2021Jan 20, 2022Apr 13, 2022Jul 5, 2022Sep 26, 2022
Updated 8mo ago

Evaluation results

MethodLinks
FAN-L-HybridParams (M)=76.8, Pre-trained=ImageNet-22K, Resolution=384x3842022.0471.1
ConvNeXt-XLData/Size=22K/384^2, FLOPs (G)=179, Params (M)=350.22022.0168.2
ConvNeXt-LData/Size=22K/384^2, FLOPs (G)=101, Params (M)=197.82022.0166.7
ConvNeXt-BData/Size=22K/384^2, FLOPs (G)=45.1, Params (M)=88.62022.0164.9
ConvNeXt-BParams (M)=88.6, Pre-trained=ImageNet-22K, Resolution=384x3842022.0464.9
FAN-L-HybridParams (M)=76.8, Pre-trained=ImageNet-22K, Resolution=224x2242022.0464.3
Swin-BParams (M)=87.8, Pre-trained=ImageNet-1K, Resolution=224x2242022.0464.2
MAEModel=ViT-L, Mixup and CutMix=w/ Mixup and CutMix2022.0960.3
GPaCoModel=ViT-L, Mixup and CutMix=w/o Mixup and CutMix2022.0960.3
MAE-VIT-LParams (M)=307, Pre-trained=ImageNet-1K, Resolution=224x2242022.0459.9
MAEModel=ViT-L, Mixup and CutMix=w/o Mixup and CutMix2022.0955.3
FAN-L-HybridParams (M)=76.8, Pre-trained=ImageNet-1K, Resolution=224x2242022.0453.2
FAN-L-VITParams (M)=80.5, Pre-trained=ImageNet-1K, Resolution=224x2242022.0453.1
FAN-B-HybridParams (M)=50.0, Pre-trained=ImageNet-1K, Resolution=224x2242022.0452.9
RVT-B*Pre-training=ImageNet-22K, Fine-tuning=ImageNet-1K2021.0552.63
DeiT-BPre-training=ImageNet-22K, Fine-tuning=ImageNet-1K2021.0552.37
FAN-B-VITParams (M)=54.0, Pre-trained=ImageNet-1K, Resolution=224x2242022.0451.8
GPaCoModel=ViT-B, Mixup and CutMix=w/o Mixup and CutMix2022.0951.7
ConvNeXt-BData/Size=1K/224^2, FLOPs (G)=15.4, Params (M)=88.62022.0151.3
ConvNeXt-BParams (M)=88.6, Pre-trained=ImageNet-1K, Resolution=224x2242022.0451.3
FAN-S-HybridParams (M)=26.0, Pre-trained=ImageNet-1K, Resolution=224x2242022.0450.7
FAN-S-ViTParams (M)=28.0, Pre-trained=ImageNet-1K, Resolution=224x2242022.0450.4
MAEModel=ViT-B, Mixup and CutMix=w/ Mixup and CutMix2022.0949.9
RVT-BPre-training=ImageNet-22K, Fine-tuning=ImageNet-1K2021.0549.67
ConvNeXt-SParams (M)=50.2, Pre-trained=ImageNet-1K, Resolution=224x2242022.0449.5
RVT-B*Data/Size=1K/224^2, FLOPs (G)=17.7, Params (M)=91.82022.0148.7
RVT-B*Params (M)=91.8, Pre-trained=ImageNet-1K, Resolution=224x2242022.0448.7
MAE-VIT-BParams (M)=86, Pre-trained=ImageNet-1K, Resolution=224x2242022.0448.3
RVT-S*Data/Size=1K/224^2, FLOPs (G)=4.7, Params (M)=23.32022.0147.7
RVT-S*Params (M)=23.3, Pre-trained=ImageNet-1K, Resolution=224x2242022.0447.7
ConvNeXt-TData/Size=1K/224^2, FLOPs (G)=4.5, Params (M)=28.62022.0147.2
ConvNeXt-TParams (M)=28.6, Pre-trained=ImageNet-1K, Resolution=224x2242022.0447.2
Swin-BData/Size=1K/224^2, FLOPs (G)=15.4, Params (M)=87.82022.0146.6
Swin-SParams (M)=50, Pre-trained=ImageNet-1K, Resolution=224x2242022.0446.6
XCIT-S12Params (M)=26.3, Pre-trained=ImageNet-1K, Resolution=224x2242022.0445.5
XCiT-S24Params (M)=47.7, Pre-trained=ImageNet-1K, Resolution=224x2242022.0445.5
MAEModel=ViT-B, Mixup and CutMix=w/o Mixup and CutMix2022.0944.9
Swin-TData/Size=1K/224^2, FLOPs (G)=4.5, Params (M)=28.32022.0141.3
Swin-TParams (M)=28.3, Pre-trained=ImageNet-1K, Resolution=224x2242022.0441.3
BYOL + Patch-basedalpha=0.052021.1041.16
GPaCoModel=ResNet-50, Mixup and CutMix=w/o Mixup and CutMix2022.0941.1
InfoMin2021.1040.68
Supervised2021.1039.76
BYOL2021.1039.53
MoCo-v2 + Patch-basedalpha=2, training setting=IFM setting2021.1038.45
MoCo-v2 + IFMepsilon=0.1, training setting=IFM setting2021.1037.23
MoCo-v2 + IFMepsilon=0.2, training setting=IFM setting2021.1037.14
MoCo-v2 + Patch-basedmemory bank size (k)=16384, alpha=32021.1036.89
MoCo-v2 + IFMepsilon=0.05, training setting=IFM setting2021.1036.79
MoCo-v2 + Patch-basedmemory bank size (k)=8192, alpha=22021.1036.31
ResNet-50Data/Size=1K/224^2, FLOPs (G)=4.1, Params (M)=25.62022.0136.1
MoCo-v2 + Patch-basedmemory bank size (k)=16384, alpha=22021.1034.78
MoCo-v2 + Texture-basedmemory bank size (k)=16384, alpha=22021.1033.36
MoCo-v2memory bank size (k)=81922021.1033.19
MoCo-v2memory bank size (k)=163842021.1032.92
CMC2021.1032.68
MoCo-v2training setting=IFM setting, memory bank size (k)=163842021.1030.38
InsDis2021.1019.6