Conventional methods of 3D object generative modeling learn volumetric predictions using deep networks with 3D convolutional operations, which are direct analogies to classical 2D ones. However, these methods are computationally wasteful in attempt to predict 3D shapes, where information is rich only on the surfaces. In this paper, we propose a novel 3D generative modeling framework to efficiently generate object shapes in the form of dense point clouds. We use 2D convolutional operations to predict the 3D structure from multiple viewpoints and jointly apply geometric reasoning with 2D projection optimization. We introduce the pseudo-renderer, a differentiable module to approximate the true rendering operation, to synthesize novel depth maps for optimization. Experimental results for single-image 3D object reconstruction tasks show that we outperforms state-of-the-art methods in terms of shape similarity and prediction density.


The slides of the talk can be downloaded here.
(Unfortunately the presentation at AAAI was not recorded)

Code and dataset

The code is hosted on GitHub (TensorFlow).
The rendered depth image dataset can be downloaded here.


AAAI 2018 paper: [ link ]

arXiv preprint: https://arxiv.org/abs/1706.07036


  title={Learning Efficient Point Cloud Generation for Dense 3D Object Reconstruction},
  author={Lin, Chen-Hsuan and Kong, Chen and Lucey, Simon},
  booktitle={AAAI Conference on Artificial Intelligence ({AAAI})},