A TOP network is used to train a variety of models, each of which uses a different combination of hyperparameters
Comparison of the PCA weights generated by the ML model (green) and the ideal ones (blue)
To generate machine learning targets, the skin is embedded in a simulated tetrahedral mesh
Three-way comparison between linear blend skinning (left), ML deformation (middle) and the ideal deformation (right). (This specific example is locally trained on the head and neck region.)
ML Deformer
Film/TV
Simulation can be used to obtain deformations that look more realistic than linear blend skinning. However, linear blend skinning is much faster. Can we learn from simulated poses to improve on linear blend skinning? This Machine Learning example shows how linear blend skinning can be improved on by learning from random poses.
This setup uses a TOP network to control all the stages of the Machine Learning: data generation, preprocessing and training. It demonstrates the use of several new Houdini 20 features that support ML, including the new ONNX SOP, the new Principal Component Analysis SOP, the new ROP Geometry Raw Output SOP, and enhancements to the Python Script TOP.
CREATED BY
COMMENTS
salar_td 5 months, 3 weeks ago |
Heeeelllllooooo
wnschnapp 3 months, 1 week ago |
Hi There,
Is Cappi included in this project? When I open the project, it seems like certain geometry are not linked. Are the remaining files available with the Hip file download? Thanks!
Rob Chapman 1 month, 1 week ago |
looking forward to more info on this and how it connects with Unreal Engines ML deformer solution for realtime playback
Please log in to leave a comment.