IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020 (Oral)
Authors: Yue Jiang, Dantong Ji, Zhizhong Han, Matthias Zwicker
Paper: http://www.cs.umd.edu/~yuejiang/papers/SDFDiff.pdf
Video: https://www.youtube.com/watch?v=l3h9JZHAOqI&t=13s
To get started:
git clone https://github.com/YueJiang-nj/CVPR2020-SDFDiff.git
1. Python3
2. CUDA10
3. Pytorch
The project has the following file layout:
README.md
multi_view_code/
bunny.sdf
dragon.sdf
code/
main.py
renderer.cpp
renderer_kernel.cu
setup.py
single_view_code/
differentiable_rendering.py
main.py
models.py
renderer.cpp
renderer_kernel.cu
setup.py
multi_view_code contains the source code for multi-view 3D reconstruction using our SDFDiff;
single_view_code contains the source code for single-view 3D reconstruction using our SDFDiff and deep learning models;
To run multi-view 3D reconstruction example, you can follow the following steps in the folder multi_view_code/code:
-
You need to run “python setup.py install” to compile our SDF differentiable renderer.
-
Once built, you can execute the bunny reconstruction example via “python main.py”