Structured uncertainty prediction networks

Dorta, Garoe, Vicente, Sara, Agapito, Lourdes, Campbell, Neill D F and Simpson, Ivor (2018) Structured uncertainty prediction networks. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, UT, USA, 18-23 June 2018. Published in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE Xplore ISSN 1063-6919 ISBN 9781538664216

[img] PDF - Published Version
Available under License All Rights Reserved.

Download (874kB)

Abstract

This paper is the first work to propose a network to predict a structured uncertainty distribution for a synthesized image. Previous approaches have been mostly limited to predicting diagonal covariance matrices. Our novel model learns to predict a full Gaussian covariance matrix for each reconstruction, which permits efficient sampling and likelihood evaluation.
We demonstrate that our model can accurately reconstruct ground truth correlated residual distributions for synthetic datasets and generate plausible high frequency samples for real face images. We also illustrate the use of these predicted covariances for structure preserving image denoising.

Item Type: Conference Proceedings
Schools and Departments: School of Engineering and Informatics > Informatics
Research Centres and Groups: Data Science Research Group
Depositing User: Ivor Simpson
Date Deposited: 24 Jan 2020 12:03
Last Modified: 24 Jan 2020 12:03
URI: http://sro.sussex.ac.uk/id/eprint/89514

View download statistics for this item

📧 Request an update