Addressing Noisy Pixels in Weakly Supervised Semantic Segmentation with Weights Assigned
Mathematics · 15.08.2024 · 12 (16) · 2520
DOI: 10.3390/math12162520 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematics, online erschienen am 15.08.2024. Verfasst von Feng Qian und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: State key laboratory major special projects of Jilin Province Science and Technology Development Plan, Guangdong Provincial University Innovation Team Project, Guangdong Province Natural Science Foundation.
Abstract (Originalsprache)
Weakly supervised semantic segmentation (WSSS) aims to segment objects without a heavy burden of dense annotations. Pseudo-masks serve as supervisory information for training segmentation models, which is crucial to the performance of segmentation models. However, the generated pseudo-masks contain significant noisy labels, which leads to poor performance of the segmentation models trained on these pseudo-masks. Few studies address this issue, as these noisy labels remain inevitable even after the pseudo-masks are improved. In this paper, we propose an uncertainty-weight transform module to mitigate the impact of noisy labels on model performance. It is noteworthy that our approach is not aimed at eliminating noisy labels but rather enhancing the robustness of the model to noisy labels. The proposed method adopts a frequency-based approach to estimate pixel uncertainty. Moreover, the uncertainty of pixels is transformed into loss weights through a set of well-designed functions. After dynamically assigning weights, the model allocates attention to each pixel in a significantly differentiated manner. Meanwhile, the impact of noisy labels on model performance is weakened. Experiments validate the effectiveness of the proposed method, achieving state-of-the-art results of 69.3% on PASCAL VOC 2012 and 39.3% on MS COCO 2014, respectively.
Förderung
- State key laboratory major special projects of Jilin Province Science and Technology Development Plan
- Guangdong Provincial University Innovation Team Project
- Guangdong Province Natural Science Foundation
- Songshan Lake Sci-tech Commissoner Program
Zitieren
APA
Qian, F., Yang, J., Tang, S., Chen, G., & Yan, J. (2024). Addressing Noisy Pixels in Weakly Supervised Semantic Segmentation with Weights Assigned. Mathematics, 12(16), 2520. https://doi.org/10.3390/math12162520
Vancouver
Qian F, Yang J, Tang S, Chen G, Yan J. Addressing Noisy Pixels in Weakly Supervised Semantic Segmentation with Weights Assigned. Mathematics. 2024;12(16):2520. doi:10.3390/math12162520
BibTeX
@article{Qian2024,
author = {Qian, Feng and Yang, Juan and Tang, Sipeng and Chen, Gao and Yan, Jingwen},
title = {Addressing Noisy Pixels in Weakly Supervised Semantic Segmentation with Weights Assigned},
journal = {Mathematics},
year = {2024},
volume = {12},
number = {16},
pages = {2520},
doi = {10.3390/math12162520}
}
EUVIMED