CLAD: A Contrastive Learning based Approach for Background Debiasing
Ke Wang (EPFL),* Harshitha Machiraju (EPFL), Oh-Hyeon Choung (Epfl), Michael Herzog (EPFL), Pascal Frossard (EPFL)
The 33rd British Machine Vision Conference
Abstract
Convolutional neural networks (CNNs) have achieved superhuman performance in multiple vision tasks, especially image classification. However, unlike humans, CNNs leverage spurious features, such as background information to make decisions. This tendency creates different problems in terms of robustness or weak generalization performance. Through our work, we introduce a contrastive learning-based approach (CLAD) to mitigate the background bias in CNNs. CLAD encourages semantic focus on object foregrounds and penalizes learning features from irrelevant backgrounds. Our method also introduces an efficient way of sampling negative samples. We achieve state-of-the-art results on the Background Challenge dataset, outperforming the previous benchmark with a margin of 4.1%. Our paper shows how CLAD serves as a proof of concept for debiasing of spurious features, such as background and texture (in supplementary material).Video
Citation
@inproceedings{Wang_2022_BMVC,
author = {Ke Wang and Harshitha Machiraju and Oh-Hyeon Choung and Michael Herzog and Pascal Frossard},
title = {CLAD: A Contrastive Learning based Approach for Background Debiasing},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0449.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by