IRUS Total

Scalable uncertainty for computer vision with functional variational inference

Title: Scalable uncertainty for computer vision with functional variational inference
Authors: Carvalho, EDC
Clark, R
Nicastro, A
Kelly, PHJ
Item Type: Conference Paper
Abstract: As Deep Learning continues to yield successful applications in Computer Vision, the ability to quantify all forms of uncertainty is a paramount requirement for its safe and reliable deployment in the real-world. In this work, we leverage the formulation of variational inference in function space, where we associate Gaussian Processes (GPs) to both Bayesian CNN priors and variational family. Since GPs are fully determined by their mean and covariance functions, we are able to obtain predictive uncertainty estimates at the cost of a single forward pass through any chosen CNN architecture and for any supervised learning task. By leveraging the structure of the induced covariance matrices, we propose numerically efficient algorithms which enable fast training in the context of high-dimensional tasks such as depth estimation and semantic segmentation. Additionally, we provide sufficient conditions for constructing regression loss functions whose probabilistic counterparts are compatible with aleatoric uncertainty quantification.
Date of Acceptance: 24-Feb-2020
URI: http://hdl.handle.net/10044/1/79364
Publisher: IEEE
Start Page: 12003
End Page: 12013
Journal / Book Title: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Copyright Statement: © 2020 The Auhtor(s). This CVPR 2020 paper is the Open Access version, provided by the Computer Vision Foundation.
Sponsor/Funder: Engineering & Physical Science Research Council (EPSRC)
Funder's Grant Number: EP/P010040/1
Conference Name: CVPR 2020
Keywords: cs.CV
Notes: CVPR 2020
Publication Status: Published online
Start Date: 2020-06-14
Finish Date: 2020-06-19
Online Publication Date: 2020-06-19
Appears in Collections:Computing
Faculty of Engineering