When working with three-dimensional data, choice of representation is key. We
explore voxel-based models, and present evidence for the viability of
voxellated representations in applications including shape modeling and object
classification. Our key contributions are methods for training voxel-based
variational autoencoders, a user interface for exploring the latent space
learned by the autoencoder, and a deep convolutional neural network
architecture for object classification. We address challenges unique to
voxel-based representations, and empirically evaluate our models on the
ModelNet benchmark, where we demonstrate a 51.5% relative improvement in the
state of the art for object classification.
Description
[1608.04236] Generative and Discriminative Voxel Modeling with Convolutional Neural Networks
%0 Generic
%1 brock2016generative
%A Brock, Andrew
%A Lim, Theodore
%A Ritchie, J. M.
%A Weston, Nick
%D 2016
%K 2016 3D GAN arxiv deep-learning paper voxel
%T Generative and Discriminative Voxel Modeling with Convolutional Neural
Networks
%U http://arxiv.org/abs/1608.04236
%X When working with three-dimensional data, choice of representation is key. We
explore voxel-based models, and present evidence for the viability of
voxellated representations in applications including shape modeling and object
classification. Our key contributions are methods for training voxel-based
variational autoencoders, a user interface for exploring the latent space
learned by the autoencoder, and a deep convolutional neural network
architecture for object classification. We address challenges unique to
voxel-based representations, and empirically evaluate our models on the
ModelNet benchmark, where we demonstrate a 51.5% relative improvement in the
state of the art for object classification.
@misc{brock2016generative,
abstract = {When working with three-dimensional data, choice of representation is key. We
explore voxel-based models, and present evidence for the viability of
voxellated representations in applications including shape modeling and object
classification. Our key contributions are methods for training voxel-based
variational autoencoders, a user interface for exploring the latent space
learned by the autoencoder, and a deep convolutional neural network
architecture for object classification. We address challenges unique to
voxel-based representations, and empirically evaluate our models on the
ModelNet benchmark, where we demonstrate a 51.5% relative improvement in the
state of the art for object classification.},
added-at = {2018-06-30T07:36:51.000+0200},
author = {Brock, Andrew and Lim, Theodore and Ritchie, J. M. and Weston, Nick},
biburl = {https://www.bibsonomy.org/bibtex/21b44c0079735fe90be67e3f397e56edf/analyst},
description = {[1608.04236] Generative and Discriminative Voxel Modeling with Convolutional Neural Networks},
interhash = {25742ac97bb5a1953b586c058d245ed7},
intrahash = {1b44c0079735fe90be67e3f397e56edf},
keywords = {2016 3D GAN arxiv deep-learning paper voxel},
note = {cite arxiv:1608.04236Comment: 9 pages, 5 figures, 2 tables},
timestamp = {2018-06-30T07:36:51.000+0200},
title = {Generative and Discriminative Voxel Modeling with Convolutional Neural
Networks},
url = {http://arxiv.org/abs/1608.04236},
year = 2016
}