Multi-stage fine-tuning of EfficientNetV2-S for material recognition on edge devices
Journal of Electrical Engineering · 01.04.2026 · 77 (2) · 123-129
DOI: 10.2478/jee-2026-0013 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Electrical Engineering, online erschienen am 18.04.2026. Verfasst von Stefan Tomić und 5 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Accurate material recognition with low computational overhead is critical for edge applications such as autonomous drones, mobile robots, and smart manufacturing systems. Direct fine-tuning of deep backbones often leads to early saturation in validation accuracy due to overfitting on small, domain-specific datasets. To address this, we propose a structured multi-phase fine-tuning strategy for EfficientNetV2-S, progressively unfreezing layers over four stages with adaptive learning rate scheduling. The approach also incorporates label smoothing, dropout, and data augmentation to enhance generalization. We evaluated the method on a curated dataset of 1,730 images across four material classes: glass, metal, paper, and plastic. The resulting model achieves a validation accuracy of 95.66%, demonstrating that the proposed pipeline effectively balances accuracy and computational efficiency, making it suitable for real-time deployment on resource-constrained edge devices.
Zitieren
APA
Tomić, S., Aldhuhoori, D., Turker, T., Nikolić, J., Perić, Z., & Zese, R. (2026). Multi-stage fine-tuning of EfficientNetV2-S for material recognition on edge devices. Journal of Electrical Engineering, 77(2), 123-129. https://doi.org/10.2478/jee-2026-0013
Vancouver
Tomić S, Aldhuhoori D, Turker T, Nikolić J, Perić Z, Zese R. Multi-stage fine-tuning of EfficientNetV2-S for material recognition on edge devices. Journal of Electrical Engineering. 2026;77(2):123-129. doi:10.2478/jee-2026-0013
BibTeX
@article{Tomi2026,
author = {Tomić, Stefan and Aldhuhoori, Dhabia and Turker, Turker and Nikolić, Jelena and Perić, Zoran and Zese, Riccardo},
title = {Multi-stage fine-tuning of EfficientNetV2-S for material recognition on edge devices},
journal = {Journal of Electrical Engineering},
year = {2026},
volume = {77},
number = {2},
pages = {123--129},
doi = {10.2478/jee-2026-0013}
}
EUVIMED