Image Classification on ImageNet (Accuracy, Latency, Energy)
1.51Latency (ms)ShiftAddViT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ShiftAddViTBackbone=PVTv2-B02023.06 | 1.51 | 70.59 | 27.13 | |
| ShiftAddViTBackbone=PVTv1-T2023.06 | 1.97 | 74.93 | 72.59 | |
| ShiftAddViTBackbone=PVTv2-B12023.06 | 2.49 | 78.49 | 85.34 | |
| ShiftAddViTBackbone=DeiT-T2023.06 | 2.94 | 72.4 | 38.21 | |
| ShiftAddViTBackbone=PVTv2-B22023.06 | 4.83 | 81.32 | 163.9 | |
| MSABackbone=DeiT-T2023.06 | 5.12 | 72.2 | 66.88 | |
| EcoformerBackbone=PVTv1-T2023.06 | 7.43 | — | 93.47 | |
| EcoformerBackbone=PVTv2-B02023.06 | 7.82 | 70.44 | 33.64 | |
| EcoformerBackbone=PVTv2-B12023.06 | 8.02 | 78.38 | 106.2 | |
| EcoformerBackbone=PVTv2-B22023.06 | 15.43 | 81.28 | 198.2 |