EUVIMEDEUVIMED entdecken
Crossref

Weeds and Crops: The Stem Emergence Point Dataset Collection

Dieudonne N. D'Arnall, Andrew R. McIntyre, Lydia Bouzar-Benlabiod

Proceedings of the Conference on Robots and Vision · 27.05.2025

DOI: 10.21428/d82e957c.3d47f637 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the Conference on Robots and Vision, online erschienen am 27.05.2025. Verfasst von Dieudonne N. D'Arnall und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Weeds present a significant threat to agriculture by competing with crops for vital resources like water, nutrients, and sunlight, ultimately reducing crop yields. Detecting weeds at an early stage is crucial for implementing effective control measures such as herbicide application, mechanical removal, or other treatments, which become less efficient and more costly as crops grow. In this paper, a dataset of 2,382 images with 1,389 images from a carrot field and 993 images from a lettuce field is presented. The visual challenge lies in accurately distinguishing weeds from crops. The paper also presents three deep neural network models for the weed detection task: a standard CNN, a hybrid CNN-LSTM, and a pure LSTM model—to assess the impact of temporal information on classification performance. Model performances are analyzed under identical training and testing conditions, evaluating trade-offs in accuracy, false detection rates, and inference speed. Our findings suggest that while the pure CNN model achieves the lowest false weed detection rate, it also has the highest false crop identification rate (13\%), whereas the pure LSTM model converges faster and provides the best overall weed classification performance.

Zitieren

APA

D'Arnall, D. N., McIntyre, A. R., & Bouzar-Benlabiod, L. (2025). Weeds and Crops: The Stem Emergence Point Dataset Collection. Proceedings of the Conference on Robots and Vision. https://doi.org/10.21428/d82e957c.3d47f637

Vancouver

D'Arnall DN, McIntyre AR, Bouzar-Benlabiod L. Weeds and Crops: The Stem Emergence Point Dataset Collection. Proceedings of the Conference on Robots and Vision. 2025. doi:10.21428/d82e957c.3d47f637

BibTeX
@article{DArnall2025,
  author = {D'Arnall, Dieudonne N. and McIntyre, Andrew R. and Bouzar-Benlabiod, Lydia},
  title = {Weeds and Crops: The Stem Emergence Point Dataset Collection},
  journal = {Proceedings of the Conference on Robots and Vision},
  year = {2025},
  doi = {10.21428/d82e957c.3d47f637}
}
Kostenlos starten →