Image Classification on ImageNet-1K 22K pre-trained (val)
87.3Top-1 AccuracySwin-L
Evaluation results
| Method | Links | |
|---|---|---|
| Swin-Limage size=384x384, #param.=197M, FLOPs=103.9G, throughput (image/s)=42.12021.03 | 87.3 | |
| Swin-Bimage size=384x384, #param.=88M, FLOPs=47.0G, throughput (image/s)=84.72021.03 | 86.4 | |
| R-152x4image size=480x480, #param.=937M, FLOPs=840.5G2021.03 | 85.4 | |
| ViT-L/16image size=384x384, #param.=307M, FLOPs=190.7G, throughput (image/s)=27.32021.03 | 85.2 | |
| Swin-Bimage size=224x224, #param.=88M, FLOPs=15.4G, throughput (image/s)=278.12021.03 | 85.2 | |
| R-101x3image size=384x384, #param.=388M, FLOPs=204.6G2021.03 | 84.4 | |
| ViT-B/16image size=384x384, #param.=86M, FLOPs=55.4G, throughput (image/s)=85.92021.03 | 84 |