Image Classification on ImageNet V2 1.0 (test)
79.2Top-1 AccuracyViT-H
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ViT-Hresolution=224, nb params (x10^6)=632.1, throughput (im/s)=112, FLOPs (x10^9)=167.4, Peak Mem (MB)=69842022.04 | 79.2 | — | |
| ViT-Lresolution=384, fine-tuned=true, nb params (x10^6)=304.8, throughput (im/s)=67, FLOPs (x10^9)=191.2, Peak Mem (MB)=128662022.04 | 79.1 | — | |
| ViT-Lresolution=224, nb params (x10^6)=304.4, throughput (im/s)=277, FLOPs (x10^9)=61.6, Peak Mem (MB)=37892022.04 | 78.6 | — | |
| ViT-Bresolution=384, fine-tuned=true, nb params (x10^6)=86.9, throughput (im/s)=190, FLOPs (x10^9)=55.5, Peak Mem (MB)=89562022.04 | 77.9 | — | |
| ConvNeXt-Lresolution=384, fine-tuned=true, nb params (x10^6)=197.8, throughput (im/s)=115, FLOPs (x10^9)=101, Peak Mem (MB)=119382022.04 | 77.7 | — | |
| ConvNeXt-XLresolution=384, fine-tuned=true, nb params (x10^6)=350.2, throughput (im/s)=80, FLOPs (x10^9)=179.0, Peak Mem (MB)=162602022.04 | 77.7 | — | |
| EfficientNetV2-XLresolution=512, nb params (x10^6)=208.1, FLOPs (x10^9)=94.02022.04 | 77 | — | |
| ConvNeXt-XLresolution=224, nb params (x10^6)=350.2, throughput (im/s)=241, FLOPs (x10^9)=60.9, Peak Mem (MB)=69512022.04 | 77 | — | |
| Swin-Lresolution=384, nb params (x10^6)=196.7, throughput (im/s)=100, FLOPs (x10^9)=103.9, Peak Mem (MB)=334562022.04 | 77 | — | |
| EfficientNetV2-Lresolution=480, fine-tuned=true, nb params (x10^6)=118.5, throughput (im/s)=179, FLOPs (x10^9)=53.0, Peak Mem (MB)=95402022.04 | 76.9 | — | |
| ConvNeXt-Bresolution=384, fine-tuned=true, nb params (x10^6)=88.6, throughput (im/s)=190, FLOPs (x10^9)=45.1, Peak Mem (MB)=78512022.04 | 76.6 | — | |
| ConvNeXt-Lresolution=224, nb params (x10^6)=197.8, throughput (im/s)=344, FLOPs (x10^9)=34.4, Peak Mem (MB)=48652022.04 | 76.6 | — | |
| ViT-Bresolution=224, nb params (x10^6)=86.6, throughput (im/s)=831, FLOPs (x10^9)=17.6, Peak Mem (MB)=20782022.04 | 76.5 | — | |
| Swin-Bresolution=384, fine-tuned=true, nb params (x10^6)=87.9, throughput (im/s)=160, FLOPs (x10^9)=47.0, Peak Mem (MB)=193852022.04 | 76.3 | — | |
| Swin-Lresolution=224, nb params (x10^6)=196.5, throughput (im/s)=337, FLOPs (x10^9)=34.5, Peak Mem (MB)=73502022.04 | 76.3 | — | |
| EfficientNetV2-Mresolution=480, fine-tuned=true, nb params (x10^6)=54.1, throughput (im/s)=312, FLOPs (x10^9)=25.0, Peak Mem (MB)=71272022.04 | 75.9 | — | |
| ConvNeXt-Bresolution=224, nb params (x10^6)=88.6, throughput (im/s)=563, FLOPs (x10^9)=15.4, Peak Mem (MB)=30292022.04 | 75.6 | — | |
| Swin-Bresolution=224, nb params (x10^6)=87.8, throughput (im/s)=532, FLOPs (x10^9)=15.4, Peak Mem (MB)=46952022.04 | 74.6 | — | |
| EfficientNetV2-Sresolution=384, fine-tuned=true, nb params (x10^6)=21.5, throughput (im/s)=874, FLOPs (x10^9)=8.5, Peak Mem (MB)=45152022.04 | 74.5 | — | |
| ViT-Sresolution=224, nb params (x10^6)=22.0, throughput (im/s)=1891, FLOPs (x10^9)=4.6, Peak Mem (MB)=9872022.04 | 73.8 | — | |
| CvT*-LS-21#Param (M)=32.1, Image Size=448x448, FLOPs (G)=34.22021.07 | 72.9 | — | |
| CvT*-LS-21#Param (M)=32.1, Image Size=384x384, FLOPs (G)=23.92021.07 | 72.5 | — | |
| CvT*-LS-21#Param (M)=32.1, Image Size=224x224, FLOPs (G)=7.9, Latency (s)=0.1222021.07 | 71.9 | — | |
| CvT-13#Param (M)=20, Image Size=384x384, FLOPs (G)=31.92021.07 | 71.9 | — | |
| CvT-21#Param (M)=32, Image Size=384x384, FLOPs (G)=452021.07 | 71.9 | — | |
| CvT*-LS-21S#Param (M)=30.1, Image Size=224x224, FLOPs (G)=11.32021.07 | 71.7 | — | |
| CvT*-LS-17#Param (M)=23.7, Image Size=224x224, FLOPs (G)=9.82021.07 | 71.6 | — | |
| CvT-21#Param (M)=32, Image Size=224x224, FLOPs (G)=10.1, Latency (s)=0.1652021.07 | 71.3 | — | |
| DeiT-B#Param (M)=86, Image Size=224x224, FLOPs (G)=17.62021.07 | 70.9 | — | |
| CvT*-LS-13#Param (M)=20.3, Image Size=224x224, FLOPs (G)=4.9, Latency (s)=0.0832021.07 | 70.5 | — | |
| CvT-13#Param (M)=20, Image Size=224x224, FLOPs (G)=6.7, Latency (s)=0.1222021.07 | 70.4 | — | |
| TPTBackbone=ViT-L/142026.01 | 70 | 16.4 | |
| C-TPTBackbone=ViT-L/142026.01 | 69.9 | 14.6 | |
| O-TPTBackbone=ViT-L/142026.01 | 69 | 10.6 | |
| SoCBackbone=ViT-L/142026.01 | 68.9 | 10.4 | |
| DeiT-S#Param (M)=22, Image Size=224x224, FLOPs (G)=4.62021.07 | 68.5 | — | |
| AWTShots=162024.07 | 68.22 | — | |
| Zero-ShotBackbone=ViT-L/142026.01 | 67.9 | 3.9 | |
| ResNet-152#Param (M)=60, Image Size=224x224, FLOPs (G)=112021.07 | 67 | — | |
| TPT + CoOpShots=162024.07 | 66.83 | — | |
| DiffTPT + CoOpShots=162024.07 | 66.8 | — | |
| ResNet-101#Param (M)=45, Image Size=224x224, FLOPs (G)=7.92021.07 | 65.7 | — | |
| PromptAlignShots=162024.07 | 65.29 | — | |
| PLOT++Shots=162024.07 | 65.07 | — | |
| TPT + CoCoOpShots=162024.07 | 64.85 | — | |
| ProGradShots=162024.07 | 64.73 | — | |
| PromptSRCShots=162024.07 | 64.35 | — | |
| CoOpShots=162024.07 | 64.2 | — | |
| KgCoOpShots=162024.07 | 64.1 | — | |
| CoCoOpShots=162024.07 | 64.07 | — | |
| MaPLeShots=162024.07 | 64.07 | — | |
| POMPShots=162024.07 | 63.32 | — | |
| ResNet-50#Param (M)=25, Image Size=224x224, FLOPs (G)=4.12021.07 | 63.3 | — | |
| DiffTPT + CoCoOpShots=162024.07 | 63.2 | — |