Abstract

This paper proposes using a sparse-structured multivari-ate Gaussian to provide a closed-form approximator for the output of probabilistic ensemble models used for dense im-age prediction tasks. This is achieved through a convolutional neural network that predicts the mean and covari-ance of the distribution, where the inverse covariance is parameterised by a sparsely structured Cholesky matrix. Similarly to distillation approaches, our single network is trained to maximise the probability of samples from pre-trained probabilistic models, in this work we use a fixed en-semble of networks. Once trained, our compact represen-tation can be used to efficiently draw spatially correlated samples from the approximated output distribution. Impor-tantly, this approach captures the uncertainty and struc-tured correlations in the predictions explicitly in a formal distribution, rather than implicitly through sampling alone. This allows direct introspection of the model, enabling vi-sualisation of the learned structure. Moreover, this formu-lation provides two further benefits: estimation of a sample probability, and the introduction of arbitrary spatial conditioning at test time. We demonstrate the merits of our approach on monocular depth estimation and show that the advantages of our approach are obtained with comparable quantitative performance.

Original languageEnglish
Title of host publicationProceedings - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
PublisherIEEE
Pages366-374
Number of pages9
ISBN (Electronic)9781665469463
DOIs
Publication statusPublished - 30 Jun 2022
Event2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022 - New Orleans, USA United States
Duration: 19 Jun 202224 Jun 2022

Publication series

NameProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
Volume2022-June
ISSN (Print)1063-6919

Conference

Conference2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
Country/TerritoryUSA United States
CityNew Orleans
Period19/06/2224/06/22

Bibliographical note

Publisher Copyright:
© 2022 IEEE.

Keywords

  • 3D from single images
  • Statistical methods

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition

Fingerprint

Dive into the research topics of 'Learning Structured Gaussians to Approximate Deep Ensembles'. Together they form a unique fingerprint.

Cite this