Image Classification on ImageNet-1K 1 (val) (Accuracy and Hardware Throughput)
83.9Top-1 AccCAS-ViT-T
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CAS-ViT-TPar (M)=21.76, Flops (M)=35972024.08 | 83.9 | 84.1 | 1,084 | 38.67 | 3.83 | |
| Swin-BPar (M)=87.7, Flops (M)=151692024.08 | 83.5 | — | 616 | — | — | |
| PVTv2-B3Par (M)=45.2, Flops (M)=69162024.08 | 83.2 | — | 561 | 3.48 | 1.18 | |
| SwiftFormer-L3Par (M)=28.48, Flops (M)=40212024.08 | 83 | — | 1,309 | 42.23 | 6.48 | |
| CAS-ViT-MPar (M)=12.42, Flops (M)=18872024.08 | 82.8 | 83 | 1,566 | 50.39 | 5.49 | |
| GC VIT-XTPar (M)=19.91, Flops (M)=26832024.08 | 82 | — | 1,072 | 8.01 | 1.75 | |
| EffiFormerV2-S2Par (M)=12.69, Flops (M)=12682024.08 | 81.6 | — | 485 | 31.36 | 7.1 | |
| MobileOne-S4Par (M)=14.8, Flops (M)=29782024.08 | 81.4 | — | 1,468 | 125.2 | 9.7 | |
| PoolFormer-S36Par (M)=32.8, Flops (M)=146202024.08 | 81.4 | — | 853 | 17.29 | 2.16 | |
| Swin-TPar (M)=28.27, Flops (M)=43722024.08 | 81.3 | — | 977 | — | — | |
| MobileViTv2-2.0Par (M)=18.4, Flops (M)=55772024.08 | 81.2 | — | 974 | 9.33 | 3.23 | |
| CAS-ViT-SPar (M)=5.76, Flops (M)=9322024.08 | 80.9 | 81.1 | 2,151 | 73.06 | 9.92 | |
| MobileViTv2-1.5Par (M)=10.56, Flops (M)=31512024.08 | 80.4 | — | 1,526 | 58.38 | 6.23 | |
| PoolFormer-S24Par (M)=21.35, Flops (M)=33942024.08 | 80.3 | — | 1,258 | 36.06 | 3.26 | |
| MobileOne-S3Par (M)=10.07, Flops (M)=19022024.08 | 80 | — | 2,899 | 652.73 | 16.16 | |
| GC VIT-XXTPar (M)=11.94, Flops (M)=19392024.08 | 79.9 | — | 951 | 41.45 | 2.76 | |
| PVT-SPar (M)=24.1, Flops (M)=36872024.08 | 79.8 | — | 933 | 40.62 | 3.08 | |
| SwiftFormer-L1Par (M)=12.05, Flops (M)=16042024.08 | 79.8 | — | 2,047 | 80.62 | 9.93 | |
| ResNeXt101Par (M)=83.46, Flops (M)=155852024.08 | 79.6 | — | 606 | 36.84 | 2.75 | |
| EffiFormerV2-S1Par (M)=6.18, Flops (M)=6662024.08 | 79 | — | 881 | 55.51 | 14.33 | |
| PVTv2-B1Par (M)=14, Flops (M)=20342024.08 | 78.7 | — | 1,385 | 21.91 | 3.41 | |
| SwiftFormer-SPar (M)=6.09, Flops (M)=9912024.08 | 78.5 | — | 2,607 | 108.98 | 12.9 | |
| EMO-5MPar (M)=5.11, Flops (M)=8832024.08 | 78.4 | — | 1,971 | 53.61 | 10.43 | |
| CAS-ViT-XSPar (M)=3.2, Flops (M)=5602024.08 | 78.3 | 78.7 | 3,251 | 115.3 | 14.63 | |
| MobileViTv2-1.0Par (M)=4.88, Flops (M)=14122024.08 | 78.1 | — | 2,429 | 91.4 | 11.52 | |
| ViT-B/16Par (M)=86.42, Flops (M)=168642024.08 | 77.9 | — | 705 | 18.27 | 0.93 | |
| EdgeViT-XSPar (M)=6.78, Flops (M)=11232024.08 | 77.5 | — | 2,044 | 85.17 | 6.77 | |
| MobileOne-S1Par (M)=4.76, Flops (M)=8302024.08 | 77.4 | — | 4,323 | 1,366.8 | 20.85 | |
| PoolFormer-S12Par (M)=11.9, Flops (M)=18132024.08 | 77.2 | — | 2,428 | 75.51 | 6.33 | |
| EfficientViT-M5Par (M)=12.47, Flops (M)=5252024.08 | 77.1 | — | 5,104 | 277.1 | — | |
| EffiFormerV2-S0Par (M)=3.59, Flops (M)=4002024.08 | 75.7 | — | 1,008 | 73.83 | 19.13 | |
| MobileViT-XSPar (M)=2.31, Flops (M)=7062024.08 | 75.7 | — | 2,893 | 129.13 | 17.9 | |
| SwiftFormer-XSPar (M)=3.47, Flops (M)=6082024.08 | 75.7 | — | 3,184 | 132.69 | 17.11 | |
| MobileNetv3-L-1.0Par (M)=4.18, Flops (M)=2152024.08 | 75.2 | — | 6,303 | 226.38 | 36.7 | |
| EMO-2MPar (M)=2.38, Flops (M)=4252024.08 | 75.1 | — | 2,672 | 80.11 | 19.07 | |
| EdgeViT-XXSPar (M)=4.07, Flops (M)=5462024.08 | 74.4 | — | 2,438 | 107.16 | 10.31 | |
| EfficientViT-M4Par (M)=8.8, Flops (M)=3012024.08 | 74.3 | — | 6,975 | 404.03 | — |