Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

SWAG: Superpixels weighted by average gradients for explanations of CNNs

Hartley, Thoma, Sidorov, Kirill ORCID: https://orcid.org/0000-0001-7935-4132, Willis, Christopher and Marshall, David ORCID: https://orcid.org/0000-0003-2789-1395 2021. SWAG: Superpixels weighted by average gradients for explanations of CNNs. Presented at: Winter Conference on Applications of Computer Vision (WACV 2021), Online / Waikoloa, HI, USA, 5-9 January 2021. Proceedings of the Winter Conference on Applications of Computer Vision. IEEE, pp. 423-432. 10.1109/WACV48630.2021.00047

[thumbnail of 0612_camera_ready.pdf]
Preview
PDF - Accepted Post-Print Version
Download (2MB) | Preview

Abstract

Providing an explanation of the operation of CNNs that is both accurate and interpretable is becoming essential in fields like medical image analysis, surveillance, and autonomous driving. In these areas, it is important to have confidence that the CNN is working as expected and explanations from saliency maps provide an efficient way of doing this. In this paper, we propose a pair of complementary contributions that improve upon the state of the art for region-based explanations in both accuracy and utility. The first is SWAG, a method for generating accurate explanations quickly using superpixels for discriminative regions which is meant to be a more accurate, efficient, and tunable drop in replacement method for Grad-CAM, LIME, or other region-based methods. The second contribution is based on an investigation into how to best generate the superpixels used to represent the features found within the image. Using SWAG, we compare using superpixels created from the image, a combination of the image and backpropagated gradients, and the gradients themselves. To the best of our knowledge, this is the first method proposed to generate explanations using superpixels explicitly created to represent the discriminative features important to the network. To compare we use both ImageNet and challenging fine-grained datasets over a range of metrics. We demonstrate experimentally that our methods provide the best local and global accuracy compared to Grad-CAM, Grad-CAM++, LIME, XRAI, and RISE.

Item Type: Conference or Workshop Item (Paper)
Date Type: Published Online
Status: Published
Schools: Professional Services > Advanced Research Computing @ Cardiff (ARCCA)
Schools > Computer Science & Informatics
Publisher: IEEE
ISBN: 9781665446402
ISSN: 2472-6737
Date of First Compliant Deposit: 15 December 2020
Date of Acceptance: 2 November 2020
Last Modified: 21 Aug 2025 13:58
URI: https://orca.cardiff.ac.uk/id/eprint/136991

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics