Learning Weakly Supervised Semantic Segmentation Through Cross-Supervision and Contrasting of Pixel-Level Pseudo-Labels

Lucas David, Helio Pedrini, Zanoni Dias

2025

Abstract

The quality of the pseudo-labels employed in training is paramount for many Weakly Supervised Semantic Segmentation techniques, which are often limited by their associated uncertainty. A common strategy found in the literature is to employ confidence thresholds to filter unreliable pixel labels, improving the overall quality of label information, but discarding a considerable amount of data. In this paper, we investigate the effectiveness of cross-supervision and contrastive learning of pixel-level pseudo-annotations in weakly supervised tasks, where only image-level annotations are available. We propose CSRM: a multi-branch deep convolutional network that leverages reliable pseudo-labels to learn to classify and segment a task in a mutual promotion scheme, while employing both reliable and unreliable pixel-level pseudo-labels to learn representations in a contrastive learning scheme. Our solution achieves 75.0% mIoU in Pascal VOC 2012 testing and 50.4% MS COCO 2014 validation datasets, respectively. Code available at github.com/lucasdavid/wsss-csrm.

Paper Citation

in Harvard Style

David L., Pedrini H. and Dias Z. (2025). Learning Weakly Supervised Semantic Segmentation Through Cross-Supervision and Contrasting of Pixel-Level Pseudo-Labels. In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 3: VISAPP; ISBN 978-989-758-728-3, SciTePress, pages 154-165. DOI: 10.5220/0013238400003912

in Bibtex Style

@conference{visapp25,
author={Lucas David and Helio Pedrini and Zanoni Dias},
title={Learning Weakly Supervised Semantic Segmentation Through Cross-Supervision and Contrasting of Pixel-Level Pseudo-Labels},
booktitle={Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 3: VISAPP},
year={2025},
pages={154-165},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013238400003912},
isbn={978-989-758-728-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 3: VISAPP
TI - Learning Weakly Supervised Semantic Segmentation Through Cross-Supervision and Contrasting of Pixel-Level Pseudo-Labels
SN - 978-989-758-728-3
AU - David L.
AU - Pedrini H.
AU - Dias Z.
PY - 2025
SP - 154
EP - 165
DO - 10.5220/0013238400003912
PB - SciTePress