pos_map prediction

[Kiord]

Hi, very nice work !

In the code there are multiple occurences of pos_map and pos_map_can as prediction types. I assume that they correspond to the object/face space 3D positions ?
The pretrained weights are only available for the uv and normal predictions though.
In the paper, I'm not sure I've read any mention of it.

Could you share some insights for that kind of predictions and their quality ?

Thanks !