P.S.:
I also found a solution to a similar problem in this thread.
In the spreadsheet, the weight paint of InnerFat is set to 0 or 1, but in the “Add Tetrahedron Collection” downstream in the graph, this value is 0.6, 0.7, etc.
Is there something wrong with this?
In the video, you set “Collection Select By Attr” to “>0.5”, which I thought was to solve this problem, but maybe not?
Creating a ground truth for ML deformers requires a long TimeRange, which is not realistic at the moment.
This is experimental and I would like to know if it is unstable depending on the environment or if it is something that should be successful as long as the operation is correct.