Image Classification on ImageNet-1k 1.0 (val) (Top-1 Accuracy, GPU Latency, Mobile Latency)
0.849Top-1 AccuracyFastViT-MA36
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FastViT-MA36Eval Image Size=384x384, Param (M)=42.7, FLOPs (G)=17.72023.03 | 0.849 | 16.5 | 9.2 | |
| NFNet-F1Eval Image Size=320x320, Param (M)=132.6, FLOPs (G)=35.52023.03 | 0.847 | 16.9 | 16.1 | |
| MOAT-2Eval Image Size=224x224, Param (M)=73.4, FLOPs (G)=17.22023.03 | 0.847 | — | 12.6 | |
| LITv2-BEval Image Size=384x384, Param (M)=87, FLOPs (G)=39.72023.03 | 0.847 | — | — | |
| MOAT-0Eval Image Size=384x384, Param (M)=27.8, FLOPs (G)=18.22023.03 | 0.846 | — | 24.3 | |
| CMT-B (GELU)Eval Image Size=256x256, Param (M)=45.7, FLOPs (G)=9.32023.03 | 0.845 | 26.9 | 20.6 | |
| Swin-BEval Image Size=384x384, Param (M)=88, FLOPs (G)=472023.03 | 0.845 | — | 57.2 | |
| MaxViT-SEval Image Size=224x224, Param (M)=69, FLOPs (G)=11.72023.03 | 0.845 | — | 48.7 | |
| FastViT-SA36Eval Image Size=384x384, Param (M)=30.4, FLOPs (G)=12.62023.03 | 0.845 | 12.8 | 7.4 | |
| EfficientNet-B7Eval Image Size=600x600, Param (M)=66, FLOPs (G)=372023.03 | 0.843 | 72.6 | 85.5 | |
| MOAT-1Eval Image Size=224x224, Param (M)=41.6, FLOPs (G)=9.12023.03 | 0.842 | — | 8.3 | |
| CoAtNet-2Eval Image Size=224x224, Param (M)=75, FLOPs (G)=15.72023.03 | 0.841 | — | 19.6 | |
| EfficientNet-B6Eval Image Size=528x528, Param (M)=43, FLOPs (G)=192023.03 | 0.84 | 44.3 | 51.7 | |
| CoAtNet-0Eval Image Size=384x384, Param (M)=25, FLOPs (G)=13.42023.03 | 0.839 | — | 20.1 | |
| EfficientNetV2-SEval Image Size=384x384, Param (M)=22, FLOPs (G)=8.82023.03 | 0.839 | 8.2 | 7.1 | |
| FastViT-MA36Eval Image Size=256x256, Param (M)=42.7, FLOPs (G)=7.92023.03 | 0.839 | 6.7 | 4.5 | |
| ConvNeXt-B (Efficient)Eval Image Size=224x224, Param (M)=89, FLOPs (G)=15.42023.03 | 0.838 | 13.6 | 8.4 | |
| MaxViT-TEval Image Size=224x224, Param (M)=31, FLOPs (G)=5.62023.03 | 0.836 | — | 31.5 | |
| LITv2-BEval Image Size=224x224, Param (M)=87, FLOPs (G)=13.22023.03 | 0.836 | — | — | |
| EfficientNet-B5Eval Image Size=456x456, Param (M)=30, FLOPs (G)=9.92023.03 | 0.836 | 24.2 | 22.1 | |
| NFNet-F0Eval Image Size=256x256, Param (M)=71.5, FLOPs (G)=12.42023.03 | 0.836 | 6.2 | 6.4 | |
| FastViT-SA36Eval Image Size=256x256, Param (M)=30.4, FLOPs (G)=5.62023.03 | 0.836 | 5.2 | 3.5 | |
| Swin-BEval Image Size=224x224, Param (M)=88, FLOPs (G)=15.42023.03 | 0.835 | — | 18.5 | |
| CMT-S (GELU)Eval Image Size=224x224, Param (M)=25.1, FLOPs (G)=42023.03 | 0.835 | 13.8 | 15.6 | |
| MOAT-0Eval Image Size=224x224, Param (M)=27.8, FLOPs (G)=5.72023.03 | 0.833 | — | 4.9 | |
| CoAtNet-1Eval Image Size=224x224, Param (M)=42, FLOPs (G)=8.42023.03 | 0.833 | — | 13.4 | |
| ConvNeXt-S (Efficient)Eval Image Size=224x224, Param (M)=50, FLOPs (G)=8.72023.03 | 0.831 | 10.3 | 5.4 | |
| ResNeSt101Eval Image Size=224x224, Param (M)=48.2, FLOPs (G)=10.22023.03 | 0.83 | 75.5 | 37.7 | |
| Swin-SEval Image Size=224x224, Param (M)=50, FLOPs (G)=8.72023.03 | 0.83 | — | 13 | |
| FastViT-SA24Eval Image Size=256x256, Param (M)=20.6, FLOPs (G)=3.82023.03 | 0.826 | 3.8 | 2.6 | |
| PoolFormer-M48Eval Image Size=224x224, Param (M)=73, FLOPs (G)=11.62023.03 | 0.825 | 30.2 | 7.6 | |
| ConvNeXt-T (Efficient)Eval Image Size=224x224, Param (M)=29, FLOPs (G)=4.52023.03 | 0.821 | 6.3 | 3.7 | |
| LITv2-SEval Image Size=224x224, Param (M)=28, FLOPs (G)=3.72023.03 | 0.82 | — | — | |
| CMT-XS (GELU)Eval Image Size=192x192, Param (M)=15.2, FLOPs (G)=1.52023.03 | 0.818 | 8.3 | 6.3 | |
| DeiT-BEval Image Size=224x224, Param (M)=86, FLOPs (G)=17.52023.03 | 0.818 | 11.6 | 11.6 | |
| RSB-ResNet152Eval Image Size=224x224, Param (M)=60.2, FLOPs (G)=11.62023.03 | 0.818 | 5.4 | 4.7 | |
| EfficientNet-B3Eval Image Size=300x300, Param (M)=12, FLOPs (G)=1.82023.03 | 0.816 | 5.3 | 7.1 | |
| CvT-13Eval Image Size=224x224, Param (M)=20.1, FLOPs (G)=4.52023.03 | 0.816 | 18.1 | 62.6 | |
| CoAtNet-0Eval Image Size=224x224, Param (M)=25, FLOPs (G)=4.22023.03 | 0.816 | — | 6.7 | |
| PoolFormer-S36Eval Image Size=224x224, Param (M)=31, FLOPs (G)=52023.03 | 0.814 | 23.7 | 3.5 | |
| Swin-TEval Image Size=224x224, Param (M)=29, FLOPs (G)=4.52023.03 | 0.813 | — | 9.3 | |
| ResNeSt50Eval Image Size=224x224, Param (M)=27.5, FLOPs (G)=5.42023.03 | 0.811 | 51.5 | 29.9 | |
| FastViT-SA12Eval Image Size=256x256, Param (M)=10.9, FLOPs (G)=1.92023.03 | 0.806 | 2.5 | 1.6 | |
| RegNetY-16GEval Image Size=224x224, Param (M)=83.6, FLOPs (G)=162023.03 | 0.804 | 8.1 | 7.5 | |
| PoolFormer-S24Eval Image Size=224x224, Param (M)=21, FLOPs (G)=3.42023.03 | 0.803 | 15.7 | 2.5 | |
| EfficientNet-B2Eval Image Size=288x288, Param (M)=9.2, FLOPs (G)=12023.03 | 0.801 | 4.1 | 3.5 | |
| RegNetY-8GEval Image Size=224x224, Param (M)=39.2, FLOPs (G)=82023.03 | 0.799 | 5.8 | 3.8 | |
| RSB-ResNet50Eval Image Size=224x224, Param (M)=25.6, FLOPs (G)=4.12023.03 | 0.798 | 2.3 | 2.6 | |
| DeiT-SEval Image Size=224x224, Param (M)=22, FLOPs (G)=4.62023.03 | 0.798 | 5.2 | 5.3 | |
| FastViT-S12Eval Image Size=256x256, Param (M)=8.8, FLOPs (G)=1.82023.03 | 0.798 | 2.2 | 1.4 | |
| MobileOne-S4Eval Image Size=224x224, Param (M)=14.8, FLOPs (G)=2.92023.03 | 0.794 | 2.6 | 1.9 | |
| CMT-T (GELU)Eval Image Size=160x160, Param (M)=9.5, FLOPs (G)=0.62023.03 | 0.791 | 4.5 | 3.8 | |
| EfficientNet-B1Eval Image Size=256x256, Param (M)=7.8, FLOPs (G)=0.72023.03 | 0.791 | 3.2 | 2.5 | |
| FastViT-T12Eval Image Size=256x256, Param (M)=6.8, FLOPs (G)=1.42023.03 | 0.791 | 2.1 | 1.2 | |
| MobileOne-S3Eval Image Size=224x224, Param (M)=10.1, FLOPs (G)=1.82023.03 | 0.781 | 2.1 | 1.5 | |
| FastViT-T8Eval Image Size=256x256, Param (M)=3.6, FLOPs (G)=0.72023.03 | 0.756 | 1.7 | 0.8 | |
| MobileOne-S0Eval Image Size=224x224, Param (M)=2.1, FLOPs (G)=0.32023.03 | 0.714 | 1 | 0.8 |