thomasodum / cvpr2018-shape-completion

CVPR 2018 paper "Learning 3D Shape Completion from Laser Scan Data with Weak Supervision".

Learning 3D Shape Completion from Laser Scan Data with Weak Supervision

Note: An updated and improved version of this approach is available as pre-print on ArXiv and the corresponding repository is davidstutz/arxiv2018-improved-shape-completion.

This repository contains paper and code corresponding to:

D. Stutz, A. Geiger. Learning 3D Shape Completion from Laser Scan Data with Weak Supervision. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.

Please cite as:

@inproceedings{Stutz2018CVPR,
  title = {Learning 3D Shape Completion from Laser Scan Data with Weak Supervision },
  author = {Stutz, David and Geiger, Andreas},
  booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  publisher = {IEEE Computer Society},
  year = {2018}
}

Also check the project page for the final publication, code and data.

Here, you can find:

• paper/: the LaTeX source of the final paper.
• code/:
  • davidstutz/daml-shape-completion, Torch and C++ implementation of the proposed approach and baselines as well as the created benchmarks.
  • davidstutz/mesh-evaluation, C++ implementation of mesh-to-mesh / mesh-to-point distance used for evaluation.
  • davidstutz/bpy-visualization-utils, Python and Blender (bpy) utilities for visualization as shown below.
  • davidstutz/mesh-voxelization, C++ implementation of mesh voxelization for computing occupancy grids and signed distance functions from watertight meshes.