MagicSource

@benikabocha thanks, it's promising. So that I can reveal almost the same function as original MMD. I have one more question, I using Saba_viewer to view a pmx model file....

@benikabocha thanks. what is physical animation? how does that information stored beside the pmx model itself?

@yuanli2333 Hopefully you guys using detectron2 rather than mmdetection.

No, I just think d2 could be simpler to follow, haha. mmdet more academic though.

@sanweiliti hi, may I ask the 3d pose point predicted from 3d pose model such VIBE or PARE can be used for any stage from LEMO?

@Khrylx hi, are u using 9.1.0?

@haofanwang I installed 0.31.3, the above test script author provided even can not rung: ``` pyvista/_vtk.py", line 318, in from vtkmodules.vtkFiltersExtraction import (vtkExtractEdges, ImportError: cannot import name 'vtkExtractEdges' from 'vtkmodules.vtkFiltersExtraction...

@sunbin1357 you have to generated your mocap data first. RFC more like a refine step to refine your rough motion cap data.

@JialeCao001 thanks for your quick reply. SipMask-mmdetection I think. Old one is also OK, but not compatible with latest pytorch version though, mainly because of C++ code in ops.

@JialeCao001 this speed was caculated only with model forward time.