EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Oriented Teacher: Exploiting Spatial Resolution Disparity to Enhance Semi-Supervised Aerial Object Detection

Yuqiu Li, Jian Xue, Ke Lu

IEEE Open Journal of Signal Processing · 2026 · 7 · 472-481

DOI: 10.1109/ojsp.2026.3685499 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of Signal Processing, 2026. Verfasst von Yuqiu Li und 2 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Science Fund for Creative Research Groups of the National Natural Science Foundation of China, Joint Key Project of the National Natural Science Foundation of China, China Electronics Technology Group Corporation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study aims to exploit the issue of spatial resolution disparity in aerial imagery for enhanced Semi-Supervised Aerial Object Detection (SSAOD). The challenge could arise when the spatial resolution of the unlabeled image deviates significantly from the resolution distribution of previously processed labeled images, potentially interfering the pseudo-labeling process. This misalignment can lead to the generation of low-quality pseudo labels, increasing the occurrence of false positives and reducing the recall rate during the sample assignment phase for the student model. To address this, we introduce a novel framework, Oriented Teacher, which incorporates a Multi-resolution Teaching Strategy (MTS). This strategy employs a combination of image transformations to adaptively adjust the spatial resolution of the unlabeled image, ensuring better alignment with labeled images and higher-quality pseudo labels generated by the teacher model. Additionally, the teacher prediction scores are utilized as soft pseudo labels, providing more informative guidance. Then our framework features a Soft VariFocal Loss (SVF) mechanism, specifically designed to minimize the impact of false positives and easy negatives within the unsupervised branch. Through extensive experiments conducted on DOTA-v1.5 and DOTA-v2.0 datasets under various evaluation protocols, we demonstrate that our proposed method achieves state-of-the-art performance.

Förderung

  • Science Fund for Creative Research Groups of the National Natural Science Foundation of China
  • Joint Key Project of the National Natural Science Foundation of China
  • China Electronics Technology Group Corporation

Zitieren

APA

Li, Y., Xue, J., & Lu, K. (2026). Oriented Teacher: Exploiting Spatial Resolution Disparity to Enhance Semi-Supervised Aerial Object Detection. IEEE Open Journal of Signal Processing, 7, 472-481. https://doi.org/10.1109/ojsp.2026.3685499

Vancouver

Li Y, Xue J, Lu K. Oriented Teacher: Exploiting Spatial Resolution Disparity to Enhance Semi-Supervised Aerial Object Detection. IEEE Open Journal of Signal Processing. 2026;7:472-481. doi:10.1109/ojsp.2026.3685499

BibTeX
@article{Li2026,
  author = {Li, Yuqiu and Xue, Jian and Lu, Ke},
  title = {Oriented Teacher: Exploiting Spatial Resolution Disparity to Enhance Semi-Supervised Aerial Object Detection},
  journal = {IEEE Open Journal of Signal Processing},
  year = {2026},
  volume = {7},
  pages = {472--481},
  doi = {10.1109/ojsp.2026.3685499}
}
Kostenlos starten →