This package provides a PyTorch module that can efficiently (1) detect and (2) penalize (self-)intersections for a triangular mesh.
Software Copyright License for non-commercial scientific research purposes. Please read carefully the terms and conditions and any accompanying documentation before you download and/or use the SMPL-X/SMPLify-X model, data and software, (the "Model & Software"), including 3D meshes, blend weights, blend shapes, textures, software, scripts, and animations. By downloading and/or using the Model & Software (including downloading, cloning, installing, and any other use of this github repository), you acknowledge that you have read these terms and conditions, understand them, and agree to be bound by them. If you do not agree with these terms and conditions, you must not download and/or use the Model & Software. Any infringement of the terms of this agreement will automatically terminate your rights under this License.
This repository provides a PyTorch wrapper around a CUDA kernel that implements the method described in Maximizing parallelism in the construction of BVHs, octrees, and k-d trees. More specifically, given an input mesh it builds a BVH tree for each one and queries it for self-intersections. Moreover, we provide a conical 3D distance field based loss for resolving the interpenetrations, as in Capturing Hands in Action using Discriminative Salient Points and Physics Simulation.
Please note that in the current implementation, for batching one needs to provide meshes with the same number of faces. Moreover, the code by default works for self-penetrations of a body mesh. The module can be used also for inter-penetrations of different meshes - for this the easiest and naive approach (without additional bookkeeping) is to fuse all meshes in a single mesh and treat inter-penetrations as self-penetrations.
Before installing anything please make sure to set the environment variable
$CUDA_SAMPLES_INC to the path that contains the header helper_math.h, which
can be found in the repo CUDA Samples repository.
To install the module run the following commands:
1. Clone this repository
git clone https://github.com/vchoutas/torch-mesh-isect
cd torch-mesh-isect2. Install the dependencies
pip install -r requirements.txt 3. Run the setup.py script
python setup.py install-
Collision Detection: Given an input mesh file, detect and plot all the collisions. Use:
python examples/detect_and_plot_collisions.py PATH_TO_MESH
-
Batch Collision resolution: Resolve self-penetrations for a batch of body models. To run use:
WEIGHT=0.001 python examples/batch_smpl_untangle.py --coll_loss_weight=$WEIGHT --model_folder=$MODEL_PARENT_FOLDER --part_segm_fn=$PATH_part_segm_fn --param_fn PKL_FN1 PKL_FN2 ... PKL_FNN
where
PKL_FN*are the filenames of the .pkl files that can be downloaded here and contain the parameters for each body model.For
batch_smpl_untangle:- To download the SMPL body model please see the directions for downloding and for specific placing in folder structure and renaming.
- The file for the
part_segm_fnargument for SMPL can be downloaded here.
The code has been tested with Python 3.6, CUDA 10.0, CuDNN 7.3 and PyTorch 1.0.
If you find this code useful in your research please cite the relevant work(s) of the following list, for detecting and penalizing mesh intersections accordingly:
@inproceedings{Karras:2012:MPC:2383795.2383801,
author = {Karras, Tero},
title = {Maximizing Parallelism in the Construction of BVHs, Octrees, and K-d Trees},
booktitle = {Proceedings of the Fourth ACM SIGGRAPH / Eurographics Conference on High-Performance Graphics},
year = {2012},
pages = {33--37},
numpages = {5},
url = {https://doi.org/10.2312/EGGH/HPG12/033-037},
doi = {10.2312/EGGH/HPG12/033-037},
publisher = {Eurographics Association}
}
@article{Tzionas:IJCV:2016, title = {Capturing Hands in Action using Discriminative Salient Points and Physics Simulation},
author = {Tzionas, Dimitrios and Ballan, Luca and Srikantha, Abhilash and Aponte, Pablo and Pollefeys, Marc and Gall, Juergen},
journal = {International Journal of Computer Vision (IJCV)},
volume = {118},
number = {2},
pages = {172--193},
month = jun,
year = {2016},
url = {https://doi.org/10.1007/s11263-016-0895-4},
month_numeric = {6}
}
This repository was originally developed for SMPL-X / SMPLify-X (CVPR 2019), you might be interested in having a look: https://smpl-x.is.tue.mpg.de.
The code of this repository was implemented by Vassilis Choutas.
For questions, please contact smplx@tue.mpg.de.
For commercial licensing (and all related questions for business applications), please contact ps-licensing@tue.mpg.de. Please note that the method for this component has been patented by NVidia and a license needs to be obtained also by them.