MinMax-CAM: Improving Focus of CAM-based Visualization Techniques in Multi-label Problems

Lucas David, Helio Pedrini, Zanoni Dias

2022

Abstract

The Class Activation Map (CAM) technique (and derivations thereof) has been broadly used in the literature to inspect the decision process of Convolutional Neural Networks (CNNs) in classification problems. However, most studies have focused on maximizing the coherence between the visualization map and the position, shape and sizes of a single object of interest, and little is known about the performance of visualization techniques in scenarios where multiple objects of different labels coexist. In this work, we conduct a series of tests that aim to evaluate the efficacy of CAM techniques over distinct multi-label sets. We find that techniques that were developed with single-label classification in mind (such as Grad-CAM, Grad-CAM++ and Score-CAM) will often produce diffuse visualization maps in multi-label scenarios, overstepping the boundaries of their explaining objects onto different labels. We propose a generalization of CAM technique, based on multi-label activation maximization/minimization to create more accurate activation maps. Finally, we present a regularization strategy that encourages sparse positive weights in the classifying layer, producing cleaner activation maps and better multi-label classification scores.

Paper Citation

in Harvard Style

David L., Pedrini H. and Dias Z. (2022). MinMax-CAM: Improving Focus of CAM-based Visualization Techniques in Multi-label Problems. In Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 4: VISAPP; ISBN 978-989-758-555-5, SciTePress, pages 106-117. DOI: 10.5220/0010807800003124

in Bibtex Style

@conference{visapp22,
author={Lucas David and Helio Pedrini and Zanoni Dias},
title={MinMax-CAM: Improving Focus of CAM-based Visualization Techniques in Multi-label Problems},
booktitle={Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 4: VISAPP},
year={2022},
pages={106-117},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010807800003124},
isbn={978-989-758-555-5},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2022) - Volume 4: VISAPP
TI - MinMax-CAM: Improving Focus of CAM-based Visualization Techniques in Multi-label Problems
SN - 978-989-758-555-5
AU - David L.
AU - Pedrini H.
AU - Dias Z.
PY - 2022
SP - 106
EP - 117
DO - 10.5220/0010807800003124
PB - SciTePress