Illumination-Guided progressive unsupervised domain adaptation for low-light instance segmentation
Neural Networks · 2025-03 · 183 · 106958
DOI: 10.1016/j.neunet.2024.106958 ↗Kurz gefasst
Fachartikel in der Zeitschrift Neural Networks, 2025-03. Verfasst von Yi Zhang und 4 weiteren Autor:innen. Beteiligte Einrichtung: School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China. Electronic address: zhangyi123@tju.edu.cn. Hauptthemen laut MEDLINE: Lighting, Unsupervised Machine Learning, Image Processing, Computer-Assisted, Neural Networks, Computer. Die Arbeit wurde bisher 11-mal zitiert (Crossref) und nennt 60 Literaturangaben. Gefördert von: National Natural Science Foundation of China, National Key Research and Development Program of China, Key Technologies Research and Development Program.
Abstract (Originalsprache)
Due to limited photons, low-light environments pose significant challenges for computer vision tasks. Unsupervised domain adaptation offers a potential solution, but struggles with domain misalignment caused by inadequate utilization of features at different stages. To address this, we propose an Illumination-Guided Progressive Unsupervised Domain Adaptation method, called IPULIS, for low-light instance segmentation by progressively exploring the alignment of features at image-, instance-, and pixel-levels between normal- and low-light conditions under illumination guidance. This is achieved through: (1) an Illumination-Guided Domain Discriminator (IGD) for image-level feature alignment using retinex-derived illumination maps, (2) a Foreground Focus Module (FFM) incorporating global information with local center features to facilitate instance-level feature alignment, and (3) a Contour-aware Domain Discriminator (CAD) for pixel-level feature alignment by matching contour vertex features from a contour-based model. By progressively deploying these modules, IPULIS achieves precise feature alignment, leading to high-quality instance segmentation. Experimental results demonstrate that our IPULIS achieves state-of-the-art performance on real-world low-light dataset LIS.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China. Electronic address: zhangyi123@tju.edu.cn.
- School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China; Shanghai Artificial Intelligence Laboratory, Shanghai, China. Electronic address: jcguo@tju.edu.cn.
- School of Physical and Mathematical Sciences, Nanyang Technological University, Singapore. Electronic address: huihui.yue@ntu.edu.sg.
- School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China. Electronic address: zhengsida@tju.edu.cn.
- School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China. Electronic address: liuchonh@tju.edu.cn.
Förderung
- National Natural Science Foundation of China
- National Key Research and Development Program of China
- Key Technologies Research and Development Program
Zitieren
APA
Zhang, Y., Guo, J., Yue, H., Zheng, S., & Liu, C. (2025). Illumination-Guided progressive unsupervised domain adaptation for low-light instance segmentation. Neural Networks, 183, 106958. https://doi.org/10.1016/j.neunet.2024.106958
Vancouver
Zhang Y, Guo J, Yue H, Zheng S, Liu C. Illumination-Guided progressive unsupervised domain adaptation for low-light instance segmentation. Neural Networks. 2025;183:106958. doi:10.1016/j.neunet.2024.106958
BibTeX
@article{Zhang2025,
author = {Zhang, Yi and Guo, Jichang and Yue, Huihui and Zheng, Sida and Liu, Chonghao},
title = {Illumination-Guided progressive unsupervised domain adaptation for low-light instance segmentation},
journal = {Neural Networks},
year = {2025},
volume = {183},
pages = {106958},
doi = {10.1016/j.neunet.2024.106958}
}
EUVIMED