Supplementary Material for Occupancy Networks : Learning 3 D Reconstruction in Function Space

In this supplementary document, we first give a detailed overview of our architectures and training procedure in Section 1. We then discuss our implementation of the baselines in Section 2 and compare them to the implementation in the original publications. Finally, we provide additional experimental results, both qualitatively and quantitatively in Section 3. The supplementary video shows 3D animations of the output of our method for the conditional tasks as well as latent space interpolations for our generative model. 1. Implementation Details In this section, we first give a detailed description of our architectures. We then describe all our data preprocessing steps and provide details on the metrics we use both for training and testing. Finally, we give details on our training protocol and inference procedure.