Fully-Convolutional Point Networks for Large-Scale Point Clouds

End-to-end learning on large-scale real-world 3D data for semantic segmentation and scene captioning.

Paper by Dario Rethage, Johanna Wald, Jürgen Sturm, Nassir Navab and Federico Tombari.

Overview

Citation

If you find our work useful, please consider citing:

  @inproceedings{rethage2018eccv,
  author  =  {Rethage, D., Wald, J. Sturm, J. Navab, N., Tombari, F.},
  title   =  {Fully-Convolutional Point Networks for Large-Scale Point Clouds},
  booktitle =  {European Conference on Computer Vision (ECCV)},
  year    =  {2018}
}

Example

a) Textured mesh (not used) b) Input point cloud c) Semantic segmentation

Note: input cloud is processed in a single pass through the network.

Code

Code and trained model coming soon!