This is pytorch type of paper《Res2Net: A New Multi-scale Backbone Architecture》，including Res2net-50, Res2Next-29。
| models | Params | FLOPs | Top-1 acc |
|---|---|---|---|
| Resnext29 8c*64w | 34.52M | 5.40G | 80.62% |
| Res2next29 6c24w4scale | 23.71M | 3.88G | 82.85% |
| Res2next29 8c25w4scale | 32.96M | 5.40G | 83.27% |
| Res2next29 6c24w6scale | 36.15M | 5.92G | 82.48% |