Occupancy Networks

@inproceedings{mescheder2019occupancy,
  title={Occupancy networks: Learning 3d reconstruction in function space},
  author={Mescheder, Lars and Oechsle, Michael and Niemeyer, Michael and
          Nowozin, Sebastian and Geiger, Andreas},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and
             Pattern Recognition},
  pages={4460--4470},
  year={2019}
}

• Related to DeepSDF but leveraging occupancy instead of distances
• Encoder-Decoder kind of training
• Can also do latent space interpolation

We then describe how we can learn a model that infers this representation from various forms of input such as point clouds, single images and low-resolution voxel representations.

Each output position is assigned a occupancy probability between 0 and 1.

Note that this network is equivalent to a neural network for binary classification, except that we are interested in the decision boundary which implicitly represents the object’s surface.

• Shape surface is then the isosurface at