UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Explaining Classifiers using Adversarial Perturbations on the Perceptual Ball

Elliott, Andrew; Law, Stephen; Russell, Chris; (2021) Explaining Classifiers using Adversarial Perturbations on the Perceptual Ball. In: Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). (pp. pp. 10688-10697). Institute of Electrical and Electronics Engineers (IEEE) Green open access

[thumbnail of Law_Explaining Classifiers using Adversarial Perturbations on the Perceptual Ball_AAM2.pdf]
Preview
Text
Law_Explaining Classifiers using Adversarial Perturbations on the Perceptual Ball_AAM2.pdf

Download (9MB) | Preview

Abstract

We present a simple regularization of adversarial perturbations based upon the perceptual loss. While the resulting perturbations remain imperceptible to the human eye, they differ from existing adversarial perturbations in that they are semi-sparse alterations that highlight objects and regions of interest while leaving the background unaltered. As a semantically meaningful adverse perturbations, it forms a bridge between counterfactual explanations and adversarial perturbations in the space of images. We evaluate our approach on several standard explainability benchmarks, namely, weak localization, insertion-deletion, and the pointing game demonstrating that perceptually regularized counterfactuals are an effective explanation for image-based classifiers.

Type: Proceedings paper
Title: Explaining Classifiers using Adversarial Perturbations on the Perceptual Ball
Event: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Location: Nashville (TN), USA
Dates: 19th-25th June 2021
ISBN-13: 978-1-6654-4509-2
Open access status: An open access version is available from UCL Discovery
DOI: 10.1109/CVPR46437.2021.01055
Publisher version: https://doi.org/10.1109/CVPR46437.2021.01055
Language: English
Additional information: This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions.
Keywords: Location awareness, Bridges, Computer vision, Perturbation methods, Neural networks, Games, Benchmark testing
UCL classification: UCL > Provost and Vice Provost Offices > UCL SLASH > Faculty of S&HS
UCL > Provost and Vice Provost Offices > UCL SLASH > Faculty of S&HS > Dept of Geography
UCL > Provost and Vice Provost Offices > UCL SLASH
UCL
URI: https://discovery.ucl.ac.uk/id/eprint/10143467
Downloads since deposit
18Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item