List your company
ResearchFormerly WizWand

Image Classification on ImageNet-1k 1.0 (test val)

86.5Top-1 AccTinyViT-21M 512
69.65274.02678.482.774Jul 21, 2022
Updated 8mo ago

Evaluation results

MethodLinks
TinyViT-21M 512#Params (M)=21, MACs (G)=27.0, Throughput (images/s)=167, Input=512, Arch.=Hybrid, Pre-training=IN-21k with fast distillation, Fine-tuning=higher resolution2022.0786.597.9
TinyViT-21M 384#Params (M)=21, MACs (G)=13.8, Throughput (images/s)=394, Input=384, Arch.=Hybrid, Pre-training=IN-21k with fast distillation, Fine-tuning=higher resolution2022.0786.297.8
TinyViT-21M#Params (M)=21, MACs (G)=4.3, Throughput (images/s)=1571, Input=224, Arch.=Hybrid, Pre-training=IN-21k with fast distillation2022.0784.897.3
EffNet-B5#Params (M)=30, MACs (G)=9.9, Throughput (images/s)=330, Input=456, Arch.=CNN2022.0783.696.7
TinyViT-11M#Params (M)=11, MACs (G)=2.0, Throughput (images/s)=2468, Input=224, Arch.=Hybrid, Pre-training=IN-21k with fast distillation2022.0783.296.5
TinyViT-21M#Params (M)=21, MACs (G)=4.3, Throughput (images/s)=1571, Input=224, Arch.=Hybrid2022.0783.196.5
CrossViT-15#Params (M)=28, MACs (G)=6.1, Throughput (images/s)=1306, Input=224, Arch.=Trans2022.0782.3—
CoaT-Lite Small#Params (M)=20, MACs (G)=4.0, Throughput (images/s)=1138, Input=224, Arch.=Trans2022.0781.995.6
AutoFormer-S#Params (M)=23, MACs (G)=5.1, Throughput (images/s)=1341, Input=224, Arch.=Trans2022.0781.795.7
LeViT-256#Params (M)=19, MACs (G)=1.1, Throughput (images/s)=7386, Input=224, Arch.=Hybrid2022.0781.6—
TinyViT-11M#Params (M)=11, MACs (G)=2.0, Throughput (images/s)=2468, Input=224, Arch.=Hybrid2022.0781.595.8
T2T-ViT-14#Params (M)=21, MACs (G)=4.8, Throughput (images/s)=1557, Input=224, Arch.=Trans2022.0781.595.7
Swin-T#Params (M)=28, MACs (G)=4.5, Throughput (images/s)=1393, Input=224, Arch.=Trans2022.0781.295.5
TinyViT-5M#Params (M)=5.4, MACs (G)=1.3, Throughput (images/s)=3060, Input=224, Arch.=Hybrid, Pre-training=IN-21k with fast distillation2022.0780.795.6
DeiT-S#Params (M)=22, MACs (G)=4.6, Throughput (images/s)=2276, Input=224, Arch.=Trans2022.0779.995
ResT-Small#Params (M)=14, MACs (G)=2.1, Throughput (images/s)=2037, Input=224, Arch.=Trans2022.0779.694.9
TinyViT-5M#Params (M)=5.4, MACs (G)=1.3, Throughput (images/s)=3060, Input=224, Arch.=Hybrid2022.0779.194.8
MobileViT-S#Params (M)=6, MACs (G)=1.8, Throughput (images/s)=2661, Input=256, Arch.=Hybrid2022.0778.4—
CrossViT-9#Params (M)=9, MACs (G)=2.0, Throughput (images/s)=2659, Input=224, Arch.=Trans2022.0777.192.4
Mobile-Former-214M#Params (M)=9, MACs (G)=0.2, Throughput (images/s)=3105, Input=224, Arch.=Hybrid2022.0776.7—
GLiT-Tiny#Params (M)=7, MACs (G)=1.5, Throughput (images/s)=3262, Input=224, Arch.=Trans2022.0776.3—
VITAS-DeiT-A#Params (M)=6, MACs (G)=1.3, Throughput (images/s)=3504, Input=224, Arch.=Trans2022.0775.592.4
PVT-Tiny#Params (M)=13, MACs (G)=1.9, Throughput (images/s)=2791, Input=224, Arch.=Trans2022.0775.1—
ResNet-18#Params (M)=12, MACs (G)=1.8, Throughput (images/s)=8714, Input=224, Arch.=CNN2022.0770.386.7