preprint · arXiv (Cornell University) · 2018

Structured Uncertainty Prediction Networks

Garoe Dorta, Sara Vicente, Lourdes Agapito, Neill D. F. Campbell, Ivor Simpson

Cited by 3Open access

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 [15]. 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.

Cite this paper

APA 7

Dorta, G., Vicente, S., Agapito, L., Campbell, N. D. F., & Simpson, I. (2018). Structured Uncertainty Prediction Networks. arXiv (Cornell University), 5477–5485. https://doi.org/10.1109/cvpr.2018.00574

MLA 9

Dorta, Garoe, et al. “Structured Uncertainty Prediction Networks.” arXiv (Cornell University), 2018, pp. 5477–85, https://doi.org/10.1109/cvpr.2018.00574.

Chicago

Dorta, Garoe, Sara Vicente, Lourdes Agapito, Neill D. F. Campbell, and Ivor Simpson. 2018. “Structured Uncertainty Prediction Networks.” arXiv (Cornell University): 5477–85. https://doi.org/10.1109/cvpr.2018.00574.

Harvard

Dorta, G. et al. (2018) ‘Structured Uncertainty Prediction Networks’, arXiv (Cornell University), pp. 5477–5485. Available at: https://doi.org/10.1109/cvpr.2018.00574.

BibTeX

@misc{dorta2018structured,
  title = {{Structured Uncertainty Prediction Networks}},
  author = {Dorta, Garoe and Vicente, Sara and Agapito, Lourdes and Campbell, Neill D. F. and Simpson, Ivor},
  journal = {arXiv (Cornell University)},
  pages = {5477--5485},
  year = {2018},
  doi = {10.1109/cvpr.2018.00574}
}

Made from the index record. Check names, capitalisation and page numbers against the paper. Need another style? AI for Research has over 10,000.