Skip to content

Latest commit

 

History

History
42 lines (34 loc) · 11.5 KB

README.md

File metadata and controls

42 lines (34 loc) · 11.5 KB
Raw

Repvgg: Making vgg-style convnets great again

Introduction

@inproceedings{ding2021repvgg,
  title={Repvgg: Making vgg-style convnets great again},
  author={Ding, Xiaohan and Zhang, Xiangyu and Ma, Ningning and Han, Jungong and Ding, Guiguang and Sun, Jian},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={13733--13742},
  year={2021}
}

Pretrain model

Model Epochs Params(M) Flops(G) Top-1 (%) Top-5 (%) Config Download
RepVGG-A0 120 9.11（train) | 8.31 (deploy) 1.52 (train) | 1.36 (deploy) 71.43 90.16 config (train) | config (deploy) model (train) | model (deploy)
RepVGG-A1 120 14.09 (train) | 12.79 (deploy) 2.64 (train) | 2.37 (deploy) 73.82 91.46 config (train) | config (deploy) model (train) |model (deploy)
RepVGG-A2 120 28.21 (train) | 25.5 (deploy) 5.7 (train) | 5.12 (deploy) 75.65 92.61 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B0 120 15.82 (train) | 14.34 (deploy) 3.42 (train) | 3.06 (deploy) 74.42 92.09 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B1 120 57.42 (train) | 51.83 (deploy) 13.16 (train) | 11.82 (deploy) 77.72 93.88 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B1g2 120 45.78 (train) | 41.36 (deploy) 9.82 (train) | 8.82 (deploy) 77.30 93.56 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B1g4 120 39.97 (train) | 36.13 (deploy) 8.15 (train) | 7.32 (deploy) 76.69 93.36 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B2 120 89.02 (train) | 80.32 (deploy) 20.46 (train) | 18.39 (deploy) 78.10 94.07 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B2g4 120 61.76 (train) | 55.78 (deploy) 12.63 (train) | 11.34 (deploy) 77.87 93.76 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B2g4 200 61.76 (train) | 55.78 (deploy) 12.63 (train) | 11.34 (deploy) 78.87 94.44 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B3 200 123.09 (train) | 110.96 (deploy) 29.17 (train) | 26.22 (deploy) 79.87 95.00 config (train) |config (deploy) model (train) |model (deploy)
RepVGG-B3g4 200 83.83 (train) | 75.63 (deploy) 17.9 (train) | 16.08 (deploy) 79.63 94.87 config (train) |config (deploy) model (train) |model (deploy)

Note: The parameters of RepVGG-D2se are not provided because the implementation of the SE module in MMClassification is different from that in the RepVGG module, resulting in the transformed RepVGG-D2se parameters not being loaded. The configuration file of RepVGG-D2se is available here.

Results and models

ImageNet

Model Params(M) Flops(G) Top-1 (%) Top-5 (%) Config Download
RepVGG-B2g4 61.76 (train) | 55.78 (deploy) 12.63 (train) | 11.34 (deploy) 76.83 93.50 config (train) | config (deploy) model (train) | model (deploy) | log
RepVGG-B3 123.09 (train) | 110.96 (deploy) 29.17 (train) | 26.22 (deploy) 77.88 (train) | 77.87 (deploy) 93.99 config (train) | config (deploy) model (train) | model (deploy) | log
RepVGG-B3g4 (training) 83.83 (train) | 75.63 (deploy) 17.9 (train) | 16.08 (deploy) config (train) | config (deploy)