EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

RADLER: Context-Aware Slimmable Model Selection for Energy-Efficient Real-Time UAV Weed Detection

Alina L. Machidon, Andraz Krasovec, Ioana C. Igret, Veljko Pejovic, Octavian M. Machidon

IEEE Access · 2026 · 14 · 116854-116866

DOI: 10.1109/access.2026.3718374 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Alina L. Machidon und 4 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Slovenian Research Agency under the project “Context-Aware On-Device Approximate Computing”, through research core funding, European Union’s Horizon 2020 research and innovation programme through the SMART4ALL project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Real-time UAV weed detection must operate under tight onboard compute and energy constraints, yet most lightweight segmentation pipelines still commit the system to a single fixed accuracy-efficiency operating point. This article introduces RADLER, a context-aware model selection strategy for slimmable neural networks that predicts, for each input image, the smallest network width that remains close to the best attainable segmentation quality. RADLER combines image-level contextual features, feature scaling with redundancy diagnostics, and threshold-controlled optimal-width labels to expose multiple operating points rather than a single adaptive configuration. We evaluate the method on two public UAV weed detection datasets, compare it against static-width baselines and an oracle reference, and characterize runtime and energy behavior on an NVIDIA Jetson Nano. Across the evaluated settings, RADLER matches full-width performance within uncertainty in one setting and otherwise trades a small, measurable IoU decrease for lower average selected width, enabling energy savings between 25% and 50% depending on the operating point. The results show that context-aware slimmable model selection can make onboard agricultural vision pipelines more transparent in their trade-offs and better suited to battery-constrained UAV platforms.

Förderung

  • Slovenian Research Agency under the project “Context-Aware On-Device Approximate Computing”
  • through research core funding
  • European Union’s Horizon 2020 research and innovation programme through the SMART4ALL project
  • 3rd Focused Technology Transfer Experiment (FTTE), “AgriAdapt.”

Zitieren

APA

Machidon, A. L., Krasovec, A., Igret, I. C., Pejovic, V., & Machidon, O. M. (2026). RADLER: Context-Aware Slimmable Model Selection for Energy-Efficient Real-Time UAV Weed Detection. IEEE Access, 14, 116854-116866. https://doi.org/10.1109/access.2026.3718374

Vancouver

Machidon AL, Krasovec A, Igret IC, Pejovic V, Machidon OM. RADLER: Context-Aware Slimmable Model Selection for Energy-Efficient Real-Time UAV Weed Detection. IEEE Access. 2026;14:116854-116866. doi:10.1109/access.2026.3718374

BibTeX
@article{Machidon2026,
  author = {Machidon, Alina L. and Krasovec, Andraz and Igret, Ioana C. and Pejovic, Veljko and Machidon, Octavian M.},
  title = {RADLER: Context-Aware Slimmable Model Selection for Energy-Efficient Real-Time UAV Weed Detection},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {116854--116866},
  doi = {10.1109/access.2026.3718374}
}
Kostenlos starten →