Differentiable volumetric rendering: Learning implicit 3D representations without 3D supervision

@InProceedings{Niemeyer_2020_CVPR,
  author = {Niemeyer, Michael and Mescheder, Lars and Oechsle, Michael and Geiger, Andreas},
  title = {Differentiable Volumetric Rendering: Learning Implicit 3D Representations Without 3D Supervision},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2020}
}

First work of learning implicit (stored in the network weights) shape and texture of scenes, which can be trained without 3D supervision (not needing 3D ground truth models).

Not bound to discretezation or low resolutions (oposed to voxel models, or meshes)
Authors design a differntiable rendering formulation.