EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Toward Versatile Small Object Detection with Temporal-YOLOv8

Martin C van Leeuwen iD, Ella P Fokkinga iD, Wyke Huizinga iD, Jan Baan, Friso G Heslinga iD

Sensors · 20.11.2024 · 24 (22) · 7387

DOI: 10.3390/s24227387 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 20.11.2024. Verfasst von Martin C van Leeuwen und 4 weiteren Autor:innen. Beteiligte Einrichtung: TNO, Defence, Safety and Security, 2597 AK The Hague, The Netherlands. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep learning has become the preferred method for automated object detection, but the accurate detection of small objects remains a challenge due to the lack of distinctive appearance features. Most deep learning-based detectors do not exploit the temporal information that is available in video, even though this context is often essential when the signal-to-noise ratio is low. In addition, model development choices, such as the loss function, are typically designed around medium-sized objects. Moreover, most datasets that are acquired for the development of small object detectors are task-specific and lack diversity, and the smallest objects are often not well annotated. In this study, we address the aforementioned challenges and create a deep learning-based pipeline for versatile small object detection. With an in-house dataset consisting of civilian and military objects, we achieve a substantial improvement in YOLOv8 (baseline mAP = 0.465) by leveraging the temporal context in video and data augmentations specifically tailored to small objects (mAP = 0.839). We also show the benefit of having a carefully curated dataset in comparison with public datasets and find that a model trained on a diverse dataset outperforms environment-specific models. Our findings indicate that small objects can be detected accurately in a wide range of environments while leveraging the speed of the YOLO architecture.

Schlagwörter

Stichwörter der Autor:innen

YOLOsmall object detectiontemporal object detection

Autor:innen und Institutionen

  • TNO, Defence, Safety and Security, 2597 AK The Hague, The Netherlands.

ORCID-Profile: Martin C van Leeuwen, Ella P Fokkinga, Wyke Huizinga, Friso G Heslinga

Zitieren

APA

van Leeuwen, M. C., Fokkinga, E. P., Huizinga, W., Baan, J., & Heslinga, F. G. (2024). Toward Versatile Small Object Detection with Temporal-YOLOv8. Sensors, 24(22), 7387. https://doi.org/10.3390/s24227387

Vancouver

van Leeuwen MC, Fokkinga EP, Huizinga W, Baan J, Heslinga FG. Toward Versatile Small Object Detection with Temporal-YOLOv8. Sensors. 2024;24(22):7387. doi:10.3390/s24227387

BibTeX
@article{vanLeeuwen2024,
  author = {van Leeuwen, Martin C and Fokkinga, Ella P and Huizinga, Wyke and Baan, Jan and Heslinga, Friso G},
  title = {Toward Versatile Small Object Detection with Temporal-YOLOv8},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {22},
  pages = {7387},
  doi = {10.3390/s24227387}
}
Kostenlos starten →