The skeletal points inferred by TorchServe do not match the examples and official website demo, and the example character skeletal cfg file loaded by fix_annotations does not match the example skeletal preview image

[TonyDua]

1.The skeletal points inferred by TorchServe do not match the examples
Inferred 👇

Example Char 1 👇

The demo version of Meta official website 👇

It can be clearly seen that four skeleton points are missing from the head of the character. The example character does not have Hip bone points. However, using the demo version of Meta official website, the deduced skeleton point position is more in line with the point position of the example role in this project. Please explain the reasons for the difference. Is there any parameter to be set for model inference? Is this pre-trained model the same as the demo version on the official website?

2.the example character skeletal cfg file loaded by fix_annotations does not match the example skeletal preview image
The fix_annotations.py load example char 1 `s cfg yaml file 👇

But the examples\characters\char1\joint_overlay.png file 👇

The demo version of Meta official website 👇

Load the CFG configuration yaml file and find that the skeleton does not match the provided preview. This is very confusing. May I ask which version shall prevail? Is it because the Flask application doesn't handle the bone point display correctly, or is the cfg configuration file a newer version and the bone preview an older version? This also doesn't match the demo on the official website.

Hope to get your guidance, thank you!