EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

CAC: Confidence-Aware Co-Training for Weakly Supervised Crack Segmentation

Fengjiao Liang iD, Qingyong Li iD, Xiaobao Li, Yang Liu, Wen Wang

Entropy · 12.04.2024 · 26 (4) · 328

DOI: 10.3390/e26040328 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Entropy, online erschienen am 12.04.2024. Verfasst von Fengjiao Liang und 4 weiteren Autor:innen. Beteiligte Einrichtung: Key Laboratory of Big Data Artificial Intelligence in Transportation (Beijing Jiaotong University), Ministry of Education, Beijing 100044, China. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 53 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Fundamental Research Funds for the Central Universities, the National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Automatic crack segmentation plays an essential role in maintaining the structural health of buildings and infrastructure. Despite the success in fully supervised crack segmentation, the costly pixel-level annotation restricts its application, leading to increased exploration in weakly supervised crack segmentation (WSCS). However, WSCS methods inevitably bring in noisy pseudo-labels, which results in large fluctuations. To address this problem, we propose a novel confidence-aware co-training (CAC) framework for WSCS. This framework aims to iteratively refine pseudo-labels, facilitating the learning of a more robust segmentation model. Specifically, a co-training mechanism is designed and constructs two collaborative networks to learn uncertain crack pixels, from easy to hard. Moreover, the dynamic division strategy is designed to divide the pseudo-labels based on the crack confidence score. Among them, the high-confidence pseudo-labels are utilized to optimize the initialization parameters for the collaborative network, while low-confidence pseudo-labels enrich the diversity of crack samples. Extensive experiments conducted on the Crack500, DeepCrack, and CFD datasets demonstrate that the proposed CAC significantly outperforms other WSCS methods.

Schlagwörter

Stichwörter der Autor:innen

co-trainingconfidence awarecrack segmentationpseudo-label dynamic divisionweakly supervised learning

Autor:innen und Institutionen

  • Key Laboratory of Big Data Artificial Intelligence in Transportation (Beijing Jiaotong University), Ministry of Education, Beijing 100044, China.
  • School of Computer Science and Technology, Jiangsu Normal University, Xuzhou 221116, China.

ORCID-Profile: Fengjiao Liang, Qingyong Li

Förderung

  • the Fundamental Research Funds for the Central Universities
  • the National Natural Science Foundation of China
  • Fundamental Research Funds for the Central Universities
  • Beijing Natural Science Foundatio
  • Shanghai Industrial Development Project

Zitieren

APA

Liang, F., Li, Q., Li, X., Liu, Y., & Wang, W. (2024). CAC: Confidence-Aware Co-Training for Weakly Supervised Crack Segmentation. Entropy, 26(4), 328. https://doi.org/10.3390/e26040328

Vancouver

Liang F, Li Q, Li X, Liu Y, Wang W. CAC: Confidence-Aware Co-Training for Weakly Supervised Crack Segmentation. Entropy. 2024;26(4):328. doi:10.3390/e26040328

BibTeX
@article{Liang2024,
  author = {Liang, Fengjiao and Li, Qingyong and Li, Xiaobao and Liu, Yang and Wang, Wen},
  title = {CAC: Confidence-Aware Co-Training for Weakly Supervised Crack Segmentation},
  journal = {Entropy},
  year = {2024},
  volume = {26},
  number = {4},
  pages = {328},
  doi = {10.3390/e26040328}
}
Kostenlos starten →