[Docs] How to provide 2D detection results for the 3D enhancement task of the Human3.6M dataset?

[chemingmei]

📚 The doc issue

I hope to use RTMPose as a 2D detector to preprocess data for the Human3.6M dataset in a 3D pose enhancement task. Currently, I directly used the pre-trained RTMDet-m RTMPose-l model as mentioned in the documentation to process the Human3.6M images, but the results are not very good. Do I need to retrain using the Human3.6M dataset? Is there a tutorial or guide for retraining, or are there other possible solutions?

Suggest a potential alternative/fix

No response