EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Diffusion-augmented YOLO26-Swin cascaded framework with hybrid SHAP-CAM for autonomous power grid inspection

Stefano Frizzo Stefenon iD, João Pedro Matos-Carvalho iD, Viviana Cocco Mariani iD, Leandro Dos Santos Coelho iD, Kin-Choong Yow iD

Autonomous Intelligent Systems · 09.06.2026 · 6 (1)

DOI: 10.1007/s43684-026-00135-2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Autonomous Intelligent Systems, online erschienen am 09.06.2026. Verfasst von Stefano Frizzo Stefenon und 4 weiteren Autor:innen. Beteiligte Einrichtung: Instituto Superior de Engenharia de Lisboa, Instituto Politécnico de Lisboa, Rua Conselheiro Emídio Navarro 1, 1959-007 Lisboa, Portugal. Die Arbeit nennt 98 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Natural Sciences and Engineering Research Council of Canada.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep learning-based autonomous inspection of power grid insulators is challenged by data imbalance and model opacity. This paper presents an end-to-end solution integrating advanced data synthesis, detection, classification, and explainability. First, a conditional diffusion model generates realistic synthetic fault images to balance the dataset. A two-stage architecture based on You Only Look Once version 26 (YOLO26) extra-large and Shifted windows (Swin)-V2-B, called YOLO26-Swin, fine-tuned with Bayesian optimization, performs robust insulator detection and then fault classification. Finally, a novel SHapley Additive exPlanations with Class Activation Mapping (SHAP-CAM) method provides intuitive visual explanations for model predictions. Extensive experiments validate our framework’s superiority: it achieves an F1-score of 0.98149 and a mean Average Precision (mAP)@[0.5] of 0.98951, exceeding leading detection and classification models. This work highlights the efficacy of diffusion models for data augmentation in critical infrastructure and advances the interpretability of vision-based inspection systems.

Schlagwörter

Stichwörter der Autor:innen

Bayesian optimizationDiffusion modelsExplainable artificial intelligenceGenerative artificial intelligence

Autor:innen und Institutionen

  • Instituto Superior de Engenharia de Lisboa, Instituto Politécnico de Lisboa, Rua Conselheiro Emídio Navarro 1, 1959-007 Lisboa, Portugal.
  • Faculty of Engineering and Applied Sciences, University of Regina, Regina, Saskatchewan S4S 0A2 Canada.
  • LASIGE, Departamento de Informática, Faculdade de Ciências, Universidade de Lisboa, 1749-016, Lisboa, Portugal.
  • COPELABS, Lusófona University, Campo Grande 376, 1749-024 Lisbon, Portugal.
  • Center of Technology and Systems (UNINOVA-CTS) and LASI, 2829-516 Caparica, Portugal.
  • Department of Electrical Engineering, Federal University of Parana, Curitiba, PR Brazil.

ORCID-Profile: Stefano Frizzo Stefenon, João Pedro Matos-Carvalho, Viviana Cocco Mariani, Leandro Dos Santos Coelho, Kin-Choong Yow

Förderung

  • Natural Sciences and Engineering Research Council of Canada

Zitieren

APA

Stefenon, S. F., Matos-Carvalho, J. P., Mariani, V. C., Dos Santos Coelho, L., & Yow, K. C. (2026). Diffusion-augmented YOLO26-Swin cascaded framework with hybrid SHAP-CAM for autonomous power grid inspection. Autonomous Intelligent Systems, 6(1). https://doi.org/10.1007/s43684-026-00135-2

Vancouver

Stefenon SF, Matos-Carvalho JP, Mariani VC, Dos Santos Coelho L, Yow KC. Diffusion-augmented YOLO26-Swin cascaded framework with hybrid SHAP-CAM for autonomous power grid inspection. Autonomous Intelligent Systems. 2026;6(1). doi:10.1007/s43684-026-00135-2

BibTeX
@article{Stefenon2026,
  author = {Stefenon, Stefano Frizzo and Matos-Carvalho, João Pedro and Mariani, Viviana Cocco and Dos Santos Coelho, Leandro and Yow, Kin-Choong},
  title = {Diffusion-augmented YOLO26-Swin cascaded framework with hybrid SHAP-CAM for autonomous power grid inspection},
  journal = {Autonomous Intelligent Systems},
  year = {2026},
  volume = {6},
  number = {1},
  doi = {10.1007/s43684-026-00135-2}
}
Kostenlos starten →