ABE-YOLO: A lightweight and efficient model for maize ear detection
Smart Agricultural Technology · 2026-08 · 14 · 102125
DOI: 10.1016/j.atech.2026.102125 ↗Kurz gefasst
Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-08. Verfasst von Zheng Duan und 10 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 58 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Key Research and Development Program of China, Chinese Academy of Agricultural Sciences, National Natural Science Foundation of China.
Abstract (Originalsprache)
Maize ear detection is critical for precision agriculture, yet remains challenging under complex field conditions, including occlusion, illumination variation, and diverse planting patterns. To address the demand for edge deployable maize ear detection, this study proposed ABE-YOLO, an enhanced version of You Only Look Once version 11 nano (YOLOv11n). The model incorporated three key improvements, namely an efficient downsampling module (ADown) for efficient feature compression with reduced spatial information loss, an enhanced feature fusion module (Conv-BiFPN) for robust multi-scale processing through learnable weighted fusion, and a lightweight detection head (EfficientHead) that reduced computational cost through grouped convolution. Among the compared models, ABE-YOLO achieved competitive accuracy (AP 50 =95.5%, AP 75 =75.8%, and AP 50-95 =65.3%) with lowered computation complexity (1.4 M parameters, 4.3 GFLOPs, and a model size of 2.9 MB) and the highest throughput (76.9 FPS on the CUDA platform). Regarding agronomy phenotyping application, ABE-YOLO achieved a center localization error of 4.29 pixels, a counting mean absolute error (MAE) of 0.386, and a mean relative error (MRE) of 11.44% for maize ear counting. External evaluation further showed that ABE-YOLO achieved the highest AP 50 of 80.8% on the unseen Guiyang dataset, with promising performance in the other detection accuracy metrics (AP 75 =24.4%, an AP 50-95 =34.2%) and counting (MAE=0.713, MRE=32.08%). These results indicated that ABE-YOLO provided a practical balance between accuracy, efficiency, and counting reliability for in-field maize ear detection and phenotypic analysis under limited computational conditions.
Förderung
- National Key Research and Development Program of China
- Chinese Academy of Agricultural Sciences
- National Natural Science Foundation of China
Zitieren
APA
Duan, Z., Jin, X., Yu, X., Ruan, X., Zhu, Y., Jiang, T., Nan, F., Yan, P., Zhang, Y., Yin, D., & Li, X. (2026). ABE-YOLO: A lightweight and efficient model for maize ear detection. Smart Agricultural Technology, 14, 102125. https://doi.org/10.1016/j.atech.2026.102125
Vancouver
Duan Z, Jin X, Yu X, Ruan X, Zhu Y, Jiang T, et al. ABE-YOLO: A lightweight and efficient model for maize ear detection. Smart Agricultural Technology. 2026;14:102125. doi:10.1016/j.atech.2026.102125
BibTeX
@article{Duan2026,
author = {Duan, Zheng and Jin, Xiuliang and Yu, Xun and Ruan, Xirui and Zhu, Yanqin and Jiang, Tiantian and Nan, Fei and Yan, Pengxiang and Zhang, Yan and Yin, Dameng and Li, Xiaojuan},
title = {ABE-YOLO: A lightweight and efficient model for maize ear detection},
journal = {Smart Agricultural Technology},
year = {2026},
volume = {14},
pages = {102125},
doi = {10.1016/j.atech.2026.102125}
}
EUVIMED