Auto-Skinning Transformations using Mixed Finite Elements Method (AST_MFEM)

Thank you for your interest in our code, and we hope you have fun trying it out! 😊

Algorithm Features:

1. Mixed Finite Element Formulation:

• Our algorithm exposes additional rotation and deformation variables in addition to vertex positions. Assuming that rotations are purely provided by the artist only at a small set of handles/joints, we formulate an optimization problem that solves for both vertex positions and symmetric part of the deformation gradient.
• Avoids specifying necessary constraints to force the simulation to track the rig motion, which can be overly constraining and produce undesirable artifacts.

2. Efficient Physics-Based Deformation:

• Ensures that the character deformations conform to the desired rig (user-defined) while avoiding performance bottlenecks typically associated with constrained physics simulations.
• Controllable skinning pipeline that can generate compelling character deformations, using a variety of physics material models.

This code can be used to re-produces the results in the following paper(s):

• Automatic Skinning using the Mixed Finite Element Method

Developed by:

• Dimitry Kachkovski
• Hongcheng Song
• Shaimaa Monem
• Abraham Kassahun Negash
• David I.W. Levin

Repository:

• https://github.com/ShMonem/AST_MFEM

License:

• Apache-2.0 see LICENSE.md.

Copyright holders:

• Developers and contributors.

Dependencies

• Use your favorite way to install dependencies from environment.yml, for example:

# Create a Conda environment from the .yml file
conda env create -f environment.yml

• Activate the virtual environment:

conda activate mfem

• We use routines from Bartels, and we include its pre-compiled binaries in \python\bartelsBin\

Code Structure

We highlight the most important parts of the structure for the user.

Item	sub-directories	data/code
data	<different_charecters>	*.mesh
		*eulers.mat
		*handles.mat
		*hierarchy.mat
		*P.mat
		*PI.mat
env		environment.yml
python	ast_fem_np	config.py
		mfem_solver_np.py
		<all_solver-functions>
		<all_skinning-functions>.py
	ui	main.py

1. In \data different meshes characters along with their skinning data are stored.
2. \env\environment.yml contains the python dependencies list to install.
3. \python\ast_fem_np is where the user can find the implementations of our solver along with all numpy based facilitating functions.
4. Simulations can be run from \ui\main.py. With minimal efforts, you can run different examples from main.py by modifying the obj_name as below:

import numpy as np
...

obj_name = 'human'
skel_anim = np.load(f'../../data/{obj_name}/{obj_name}_skel_anim.npy')