Implementation of the SDF Loss

In this package we provide a CUDA implementation of the SDF computation as well as of the SDF-based collision function.

Installation instructions

You can install the package by running the setup script. Before doing so verify the compiler, the gcc and the g++ versions!:

python setup.py install

GIVEN that the TORCH version I am using is way newer than torch 1.1 used in this CRMH (Jiang et.al) code: You have to change the C code: check https://github.com/daniilidis-group/neural_renderer/pull/86/files

change AT_CHECK for TORCH_CHECK

Acknowledgements

Our SDF CUDA kernels are based on the original C++ implementation by David Stutz.

Citing

If you find this code useful for your research or the use data generated by our method, please consider citing the following paper:

@Inproceedings{jiang2020mpshape,
Title          = {Coherent Reconstruction of Multiple Humans from a Single Image},
Author         = {Jiang, Wen and Kolotouros, Nikos and Pavlakos, Georgios and Zhou, Xiaowei and Daniilidis, Kostas},
Booktitle      = {CVPR},
Year           = {2020}
}

The original instructions are found here