Animations Snapshots Reduction Subspaces

This code generates reduced subspaces for physics-based simulations, including real-time deformable character animation.

Algorithm Features:

• Uses a snapshots method to capture the range of deformations for a character.
  • Snapshots: Collected during a range-of-motion test for a character mesh.
• Provides the closest approximate solution with a user-defined size.
• Tested on projective dynamics simulations, which use an iterative solver for position state computation.
• Faster, more stable, and visually accurate compared to Linear-blend skinning subspaces.

For more details, refer to the following publication(s):

• Improved-Projective-Dynamics-Global-Using-Snapshots-based-Reduced-Bases SIGGRAPH23 1st place student competition award-winning paper.
• On DEIM Compatibility for Constraint Projection Reduction soon available

Developed by:

• Shaimaa Monem
• Max Planck Institute for Dynamics of Complex Technical Systems, Magdeburg, Germany.

Repository:

• https://github.com/ShMonem/animSnapBases
• Moved from: https://github.com/ShMonem/Snapshots-Reduction-Subspaces-for-Projective-Dynamics

License:

• Apache-2.0 see LICENSE.md.
• This code borrwes some functions from the beautiful implementations of SPLOCS, which was published under MIT license.

Copyright holders:

• Shaimaa Monem.

Dependencies

• Install dependencies from venv_requirements.txt:

pip install -r venv_requirements.txt 

• Activate the virtual environment:
  • On Linux:

    source re_pd/bin/activate
    

  • On Windows

    re_pd/Scripts/Activate
    

Code Structure

Thank you for your interest in our code, which you all the fun trying it out :-). !The code id structured into main directories

Item	sub-directories	data/code
input_data	<mesh>\<experiment>\FOM_snapshots	*.off/*.ply
snapbases		posSnapshots.py
		posComponents.py
		nonlinear_snapshots.py
		constraintsComponents.py
projective_dynamics		demos
		main.py
		Simulators.py
		usr_interface.json
utils		utils.py
		process.py
		support.py
test		compare_npy_files.py
config		config.py
		bases_config.json

1. input_data stores your snapshots or frames in .off or .ply format. If you use a different example or file format, update the inputs in config/bases_config.json.

2. To run the provided example, execute main.py in the root directory:

   python main.py
   

3. In main.py:

• Snapshots/frames are imported, creating a .h5 file with aligned frames.
• bases container is initialized by calling the posComponents class.
• posComponents has an posSnapshots attribute that access configurations from .json.

• bases = posComponents()
  

• Compute the bases using: bases.compute_components_store_singvalues(vertPos_output_directory)
• Store the bases/components as .bin or .npy:

  bases.store_components_to_files(vertPos_output_directory, start, vertPos_numComponents, step, '.bin')
  

Reproducibility

For vertex positions reduction, a repository redPD is provided in order to test and reproduce results from the mentioned paper.

• Kindly refer to the dedicated README.test.md. It explains all steps starting from snapshots collections to bases testing.

For constarints projection reduction, the directory projective_dynamics/ contains README.pd.md that explains all steps to collect nonlinear forces snapshots and basis computations, as well as running reduced simulations.

1. Clone the repo.
2. Install the virtual environment.
3. Run the code from the root directory using python main.py with splocs->"compute": "No" and then with splocs->"compute": "Yes":

   "splocs":
   {
     "compute": "No",
   }

4. PCA--> compute is always set to Yes, otherwise no computations run.
5. Bases matrices in .bin/.npy/.h5formats, and a.csv for singular values if desired, will be stored in results/<mesh>/<experiment>/q_bases/.
6. After bases storage, both snapshots and bases can be visualized, this option can be modified in the bases_config.json. Associated animation files can be found in both results/<mesh>/<experiment>/q_snapshts_h5/ and results/<mesh>/<experiment>/q_bases/. Note: In order to stop the snapshots animations from the animations cycle, press stop animation --> OK then close the tap, otherwise it might produce an error.

For code navigation and further development, refer to config/README.dev.md.