List your company
ResearchFormerly WizWand

Image Classification on ImageNet-1k 1.0 (val) (Accuracy, Latency, Throughput)

0.841Top-1 AccConvNext-T
0.680840.722420.7640.80558Nov 22, 2021Dec 24, 2021Jan 26, 2022Feb 28, 2022Apr 1, 2022May 4, 2022Jun 6, 2022
Updated 8mo ago

Evaluation results

MethodLinks
ConvNext-TType=CNN, Neural search=false, Extra data=ImageNet-21k, Image size=384x384, Params=28.6M, FLOPs=13.1G2022.060.8410.0086645
MobileViTv2-2.0Type=Hybrid, Neural search=false, Extra data=ImageNet-21k-P, Image size=384x384, Params=18.5M, FLOPs=16.1G2022.060.8340.017488
ConvNext-TType=CNN, Neural search=false, Extra data=ImageNet-21k, Image size=224x224, Params=28.6M, FLOPs=4.5G2022.060.8290.00371,800
PoolFormer-M48General Arch.=MetaFormer, Token Mixer=Pooling, Image Size=224, Params (M)=73, MACs (G)=11.62021.110.825——
MobileViTv2-2.0Type=Hybrid, Neural search=false, Extra data=ImageNet-21k-P, Image size=256x256, Params=18.5M, FLOPs=7.5G2022.060.8240.00751,105
ConvNext-TType=CNN, Neural search=false, Extra data=None, Image size=224x224, Params=28.6M, FLOPs=4.5G2022.060.8210.00371,800
PoolFormer-M36General Arch.=MetaFormer, Token Mixer=Pooling, Image Size=224, Params (M)=56, MACs (G)=8.82021.110.821——
DeiT-BaseType=Transformer, Neural search=false, Extra data=None, Image size=224x224, Params=86.6M, FLOPs=17.6G2022.060.8180.0132958
RSB-ResNet-152General Arch.=Convolutional Neural Networks, Token Mixer=-, Image Size=224, Params (M)=60, MACs (G)=11.62021.110.818——
DeiT-BGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=86, MACs (G)=17.52021.110.818——
PVT-LargeGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=61, MACs (G)=9.82021.110.817——
gMLP-BGeneral Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=73, MACs (G)=15.82021.110.816——
PoolFormer-S36General Arch.=MetaFormer, Token Mixer=Pooling, Image Size=224, Params (M)=31, MACs (G)=52021.110.814——
Swin-TType=Hybrid, Neural search=false, Extra data=None, Image size=224x224, Params=28.3M, FLOPs=4.5G2022.060.813—1,390
RSB-ResNet-101General Arch.=Convolutional Neural Networks, Token Mixer=-, Image Size=224, Params (M)=45, MACs (G)=7.92021.110.813——
Swin-Mixer-B/D24General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=61, MACs (G)=10.42021.110.813——
MobileViTv2-2.0Type=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=18.5M, FLOPs=7.5G2022.060.8120.00751,105
PVT-MediumGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=44, MACs (G)=6.72021.110.812——
ResMLP-B24General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=116, MACs (G)=232021.110.81——
MobileViTv2-1.5Type=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=10.6M, FLOPs=4.0G2022.060.8040.00511,418
PoolFormer-S24General Arch.=MetaFormer, Token Mixer=Pooling, Image Size=224, Params (M)=21, MACs (G)=3.42021.110.803——
EfficientNet-b2Type=CNN, Neural search=true, Extra data=None, Image size=288x288, Params=9.1M, FLOPs=1.2G2022.060.8010.00382,032
RSB-ResNet-50General Arch.=Convolutional Neural Networks, Token Mixer=-, Image Size=224, Params (M)=26, MACs (G)=4.12021.110.798——
DeiT-SGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=22, MACs (G)=4.62021.110.798——
PVT-SmallGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=25, MACs (G)=3.82021.110.798——
ViT-B/16*General Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=86, MACs (G)=17.62021.110.797——
Swin-Mixer-T/D6General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=256, Params (M)=23, MACs (G)=42021.110.797——
gMLP-SGeneral Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=20, MACs (G)=4.52021.110.796——
ResMLP-S24General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=30, MACs (G)=62021.110.794——
Swin-Mixer-T/D24General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=256, Params (M)=20, MACs (G)=42021.110.794——
MobileViT-SType=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=5.6M, FLOPs=2.0G2022.060.7840.00341,986
MobileViTv2-1.0Type=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=4.9M, FLOPs=1.8G2022.060.7810.00342,351
MobileFormer-294Type=Hybrid, Neural search=false, Extra data=None, Image size=224x224, Params=11.8M, FLOPs=294M2022.060.7790.04071,402
PoolFormer-S12General Arch.=MetaFormer, Token Mixer=Pooling, Image Size=224, Params (M)=12, MACs (G)=1.82021.110.772——
EfficientNet-b0Type=CNN, Neural search=true, Extra data=None, Image size=224x224, Params=5.3M, FLOPs=422M2022.060.7710.00164,619
ResMLP-S12General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=15, MACs (G)=32021.110.766——
MLP-Mixer-B/16General Arch.=MetaFormer, Token Mixer=Spatial MLP, Image Size=224, Params (M)=59, MACs (G)=12.72021.110.764——
ViT-L/16*General Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=307, MACs (G)=63.62021.110.761——
RSB-ResNet-34General Arch.=Convolutional Neural Networks, Token Mixer=-, Image Size=224, Params (M)=22, MACs (G)=3.72021.110.755——
PVT-TinyGeneral Arch.=Attention, Token Mixer=Attention, Image Size=224, Params (M)=13, MACs (G)=1.92021.110.751——
DeiT-TinyType=Transformer, Neural search=false, Extra data=None, Image size=224x224, Params=5.5M, FLOPs=1.3G2022.060.7220.00344,541
RSB-ResNet-18General Arch.=Convolutional Neural Networks, Token Mixer=-, Image Size=224, Params (M)=12, MACs (G)=1.82021.110.706——
MobileViTv2-0.5Type=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=1.4M, FLOPs=0.5G2022.060.7020.00164,595
MobileViT-XXSType=Hybrid, Neural search=false, Extra data=None, Image size=256x256, Params=1.3M, FLOPs=0.4G2022.060.690.00484,225
MobileFormer-52Type=Hybrid, Neural search=false, Extra data=None, Image size=224x224, Params=3.6M, FLOPs=52M2022.060.6870.00714,445