Li Bin

Training multi-class network based on DetectNet is not very easy, recently I got some better results. I will share some tips on how to train this type network. 

@CPFLAME 使用coco训练也会出现类似问题

centernet_res18_coco_0.5.yaml 跑10次大概有2~3次是0.0000的mAP，什么也没改

@zc-tx `pip install -U 'git+https://github.com/CPFLAME/detectron2.git'` in https://github.com/CPFLAME/centerX/blob/master/README.md#requirements

> I have tried this new DCNv2 which you provided, but still get the same error > > > > Did you found a solution @starsky68 ? I get the...

https://github.com/lbin/DCNv2/blob/master/dcn_v2.py#L43 maybe you should clean you dcn env and recompile the dcn codes @JesseYang

@GreysonPhoenix https://github.com/lbin/DCNv2/tree/pytorch_1.6

This is the problem: mxnet #2585, pls check https://github.com/dmlc/mxnet/pull/2585

> DCNv2的pytorch-2.0.1分支，对应的pytorch版本也是2.0.1吗？ 没做具体测试