SSOD for YOLOv5

[BowieHsu]

Hi，thank you to the friends at the Hikvision Research Institute for open-sourceing your paper source code. we have implemented a semi-supervised object dection for YOLOv5, and in the design of the algorithm, we have studied your paper and code in depth, helping us to save a lot of detours. Our project address is: https://github.com/AlibabaResearch/efficientteacher.
We hope you can provide some comments and suggestions on our technical routes and code implementation, to provide a more stable semi-supervised training implementation for the community and industry.