Recent advances in geometric deep-learning introduce complex computational
challenges for evaluating the distance between meshes. From a mesh model, point
clouds are necessary along with a robust distance metric to assess surface
quality or as part of the loss function for training models. Current methods
often rely on a uniform random mesh discretization, which yields irregular
sampling and noisy distance estimation. In this paper we introduce MongeNet, a
fast and optimal transport based sampler that allows for an accurate
discretization of a mesh with better approximation properties. We compare our
method to the ubiquitous random uniform sampling and show that the
approximation error is almost half with a very small computational overhead.
Description
[2104.14554] MongeNet: Efficient Sampler for Geometric Deep Learning
%0 Generic
%1 lebrat2021mongenet
%A Lebrat, Léo
%A Cruz, Rodrigo Santa
%A Fookes, Clinton
%A Salvado, Olivier
%D 2021
%K 2021 deep-learning geometry sampling
%T MongeNet: Efficient Sampler for Geometric Deep Learning
%U http://arxiv.org/abs/2104.14554
%X Recent advances in geometric deep-learning introduce complex computational
challenges for evaluating the distance between meshes. From a mesh model, point
clouds are necessary along with a robust distance metric to assess surface
quality or as part of the loss function for training models. Current methods
often rely on a uniform random mesh discretization, which yields irregular
sampling and noisy distance estimation. In this paper we introduce MongeNet, a
fast and optimal transport based sampler that allows for an accurate
discretization of a mesh with better approximation properties. We compare our
method to the ubiquitous random uniform sampling and show that the
approximation error is almost half with a very small computational overhead.
@misc{lebrat2021mongenet,
abstract = {Recent advances in geometric deep-learning introduce complex computational
challenges for evaluating the distance between meshes. From a mesh model, point
clouds are necessary along with a robust distance metric to assess surface
quality or as part of the loss function for training models. Current methods
often rely on a uniform random mesh discretization, which yields irregular
sampling and noisy distance estimation. In this paper we introduce MongeNet, a
fast and optimal transport based sampler that allows for an accurate
discretization of a mesh with better approximation properties. We compare our
method to the ubiquitous random uniform sampling and show that the
approximation error is almost half with a very small computational overhead.},
added-at = {2021-05-06T14:43:25.000+0200},
author = {Lebrat, Léo and Cruz, Rodrigo Santa and Fookes, Clinton and Salvado, Olivier},
biburl = {https://www.bibsonomy.org/bibtex/20475510902152b7e0e3ffd9333ab9a20/analyst},
description = {[2104.14554] MongeNet: Efficient Sampler for Geometric Deep Learning},
interhash = {ab76c6069b7d5fda963ed34654d4f7d0},
intrahash = {0475510902152b7e0e3ffd9333ab9a20},
keywords = {2021 deep-learning geometry sampling},
note = {cite arxiv:2104.14554},
timestamp = {2021-05-06T14:43:25.000+0200},
title = {MongeNet: Efficient Sampler for Geometric Deep Learning},
url = {http://arxiv.org/abs/2104.14554},
year = 2021
}