Mesopotamia

The training and initialization setting for baseline and distillation should be the same. Such as using pretrained Res-50. Normally, the perfromance after the first epoch would be much higher than...

1. Not this, you do not need to skip it. 2. Keep the same, including optimizer.

It seems strange. Does the baseline keep the same that the first epoch performs best? Does the learning rate for distillation keep the same as baseline.

DWPose always predicts all the 133 key points together via a single model. You can add your own py file that just draws the hands or face.

The problem actually exists. Later we will relase a stronger DWPose model. The accuracy will gets higher.

@ZechengLi19 You can refer this [RTMW-x](https://github.com/open-mmlab/mmpose/tree/main/projects/rtmpose). Besides, you can also use it as a teacher model to train a dwpose student for better peformance.

Yes, we also notice the situation. Later we will consider training a new ControlNet model for DWPose.

I follow your code and try openpose_e2e_test.py. But I do not meet the problem. The problem is [here](https://github.com/huchenlei/sd-webui-controlnet/blob/dw_pose/annotator/openpose/wholebody.py#L97) when the point is None. After I fix this, I get the...

Maybe the pth is broken? Or there are some problems with the file dir?

@huchenlei We will release a new version with onnx. It will not rely on any mmxx.