Image Classification on ImageNet-1k 1.0 (val) (Top-1/Top-5 Accuracy, Top-1 Loss)
83Top-1 AccuracySwin-S
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Swin-SFLOPs (G)=8.7, Params (M)=49.62021.11 | 83 | — | 96.2 | |
| SPViT-Swin-SFLOPs (G)=6.1, Params (M)=38.9, fine-tuning with knowledge distillation=true2021.11 | 83 | 0 | 96.4 | |
| SPViT-DeiT-BFLOPs (G)=8.4, Params (M)=41.6, fine-tuning with knowledge distillation=true2021.11 | 82.4 | -0.6 | 96.1 | |
| SPViT-Swin-SFLOPs (G)=6.1, Params (M)=38.92021.11 | 82.4 | 0.6 | 96 | |
| MDC-DeiT-BFLOPs (G)=11.22021.11 | 82.3 | -0.5 | — | |
| S2ViTE-DeiT-BFLOPs (G)=11.7, Params (M)=56.8, sparse training from scratch=true2021.11 | 82.2 | -0.4 | — | |
| EVIT-DeiT-BFLOPs (G)=11.6, Params (M)=86.42021.11 | 82.1 | -0.3 | 95.6 | |
| DeiT-BFLOPs (G)=17.5, Params (M)=86.42021.11 | 81.8 | — | 95.6 | |
| SPViT-DeiT-BFLOPs (G)=8.4, Params (M)=41.62021.11 | 81.5 | 0.3 | 95.7 | |
| DynamicViT-DeiT-BFLOPs (G)=11.4, Params (M)=87.12021.11 | 81.4 | 0.4 | 95.5 | |
| VTP-DeiT-BFLOPs (G)=13.8, Params (M)=67.32021.11 | 81.3 | 0.5 | 95.3 | |
| IA-RED2-DeiT-BFLOPs (G)=11.8, Params (M)=86.42021.11 | 81.3 | 0.5 | — | |
| dTPS-DeiT-BFLOPs (G)=11.4, Params (M)=872021.11 | 81.2 | 0.6 | — | |
| Swin-TiFLOPs (G)=4.5, Params (M)=28.32021.11 | 81.2 | — | 95.5 | |
| eTPS-DeiT-BFLOPs (G)=11.4, Params (M)=86.42021.11 | 81.1 | 0.7 | — | |
| SPViT-Swin-TiFLOPs (G)=3.4, Params (M)=25.8, fine-tuning with knowledge distillation=true2021.11 | 81 | 0.2 | 95.4 | |
| VTP-DeiT-BFLOPs (G)=10, Params (M)=482021.11 | 80.7 | 1.1 | — | |
| UVC-DeiT-BFLOPs (G)=8, fine-tuning with knowledge distillation=true2021.11 | 80.6 | 1.2 | — | |
| SPViT-DeiT-SFLOPs (G)=3.3, Params (M)=15.9, fine-tuning with knowledge distillation=true2021.11 | 80.3 | -0.4 | 95.1 | |
| dTPS-DeiT-SFLOPs (G)=3, Params (M)=22.82021.11 | 80.1 | -0.2 | — | |
| SSP-DeiT-BFLOPs (G)=11.7, Params (M)=56.8, sparse training from scratch=true2021.11 | 80.1 | 1.7 | — | |
| SPViT-Swin-TiFLOPs (G)=3.4, Params (M)=25.82021.11 | 80.1 | 1.1 | 95 | |
| DeiT-SFLOPs (G)=4.6, Params (M)=22.12021.11 | 79.9 | — | 95 | |
| MDC-DeiT-SFLOPs (G)=2.92021.11 | 79.9 | 0 | — | |
| eTPS-DeiT-SFLOPs (G)=3, Params (M)=22.12021.11 | 79.7 | 0.2 | — | |
| STEP-Swin-SFLOPs (G)=6.3, Params (M)=36.92021.11 | 79.6 | 3.4 | 94.7 | |
| EVIT-DeiT-SFLOPs (G)=3, Params (M)=22.12021.11 | 79.5 | 0.4 | 94.8 | |
| ToMe-DeiT-SFLOPs (G)=2.7, Params (M)=22.12021.11 | 79.4 | 0.5 | — | |
| UVC-DeiT-SFLOPs (G)=2.7, fine-tuning with knowledge distillation=true2021.11 | 79.4 | 0.5 | — | |
| DynamicViT-DeiT-SFLOPs (G)=3, Params (M)=22.82021.11 | 79.3 | 0.6 | 94.7 | |
| S2ViTE-DeiT-SFLOPs (G)=3.2, Params (M)=14.6, sparse training from scratch=true2021.11 | 79.2 | 0.7 | — | |
| SPViT-DeiT-SFLOPs (G)=3.3, Params (M)=15.92021.11 | 78.3 | 1.6 | 94.3 | |
| SSP-DeiT-SFLOPs (G)=3.2, Params (M)=14.6, sparse training from scratch=true2021.11 | 77.7 | 2.2 | — | |
| STEP-Swin-TiFLOPs (G)=3.5, Params (M)=23.62021.11 | 77.2 | 4 | 93.6 | |
| SPViT-DeiT-TiFLOPs (G)=1, Params (M)=4.8, fine-tuning with knowledge distillation=true2021.11 | 73.2 | -1 | 91.4 | |
| DeiT-TiFLOPs (G)=1.3, Params (M)=5.72021.11 | 72.2 | — | 91.1 | |
| UVC-DeiT-TiFLOPs (G)=0.7, fine-tuning with knowledge distillation=true2021.11 | 71.8 | 0.4 | — | |
| SPViT-DeiT-TiFLOPs (G)=1, Params (M)=4.82021.11 | 70.7 | 1.5 | 90.3 | |
| S2ViTE-DeiT-TiFLOPs (G)=1, Params (M)=4.2, sparse training from scratch=true2021.11 | 70.1 | 2.1 | — | |
| SSP-DeiT-TiFLOPs (G)=1, Params (M)=4.2, sparse training from scratch=true2021.11 | 68.6 | 3.6 | — |