Image Classification on ImageNet-1K (val) (GPU Throughput and Multi-Hardware Latency)
1GPU Latency (ms)MobileOne-S3
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MobileOne-S3MACs (M)=18952026.03 | 1 | 2,433 | 11.8 | 1.2 | 78.1 | |
| LowFormer-E1MACs (M)=13502026.03 | 1 | 6,337 | 6.2 | 1.7 | 78.8 | |
| LowFormer-E2MACs (M)=38002026.03 | 1.5 | 2,070 | 14.7 | 2.5 | 81.6 | |
| FastViT-SA12MACs (M)=19432026.03 | 1.7 | 1,075 | 17.5 | 1.6 | 80.6 | |
| EfficientMod-xxsMACs (M)=5792026.03 | 2.1 | 2,857 | 15 | 1.71 | 76 | |
| EfficientMod-xsMACs (M)=7732026.03 | 2.5 | 2,352 | 17.7 | 2.2 | 78.3 | |
| EdgeViT-XSMACs (M)=11272026.03 | 2.7 | 2,127 | 16 | 1.5 | 77.5 | |
| GraphLeapModel Target=ViG-B / Py-B, Platform=Alveo U2802026.04 | 2.77 | — | — | — | 80.66 | |
| LowFormer-B0MACs (M)=9442026.03 | 2.9 | 5,988 | 8.5 | 1.5 | 78.4 | |
| FastViT-SA24MACs (M)=37692026.03 | 3 | 606 | 30.9 | 2.6 | 82.6 | |
| LowFormer-E3MACs (M)=53502026.03 | 3.6 | 1,566 | 25 | 3.6 | 83 | |
| EfficientMod-sMACs (M)=14022026.03 | 3.8 | 1,381 | 30.5 | 2.6 | 81 | |
| RepViT-M1.5MACs (M)=22762026.03 | 4.2 | 1,146 | 23 | 1.5 | 81.2 | |
| LowFormer-B1.5MACs (M)=25732026.03 | 4.8 | 2,739 | 18.1 | 2.8 | 81.2 | |
| LowFormer-B2MACs (M)=36892026.03 | 4.8 | 2,227 | 21.6 | 3.5 | 81.6 | |
| RepViT-M2.3MACs (M)=45202026.03 | 5.5 | 642 | 40.6 | 2.4 | 82.5 | |
| LowFormer-B3r192MACs (M)=44792026.03 | 5.5 | 1,562 | 30 | 4.5 | 82.7 | |
| UbiMoEModel Target=M3ViT (MoE), Platform=Alveo U2802026.04 | 13.6 | — | — | — | 81.4 | |
| DRViTModel Target=ViT-Base, Platform=Alveo U2502026.04 | 14.4 | — | — | — | 81.8 |