This study constructed four datasets dedicated to network development, robustness verification, transferability assessment, and phenotypic trait extraction, respectively. A real wheat dataset (RealWheat2021) was constructed for network development and performance evaluation. Three datasets were constructed to investigate OSNet's transferability to new breeding scenarios, including RealWheat2023, RealRice2023, and 3DPointWheat.
The study area is located at the Baima Experimental Station (119°18′71″E, 31°62′00″N) of Nanjing Agricultural University in China. A total of 960 plots were cultivated during the period spanning 2020 to 2021. The experimental design involved two replications, adhering to the principles of the split-plot design. Each plot occupied an area of 1.5 m × 1.5 m and consisted of six rows of wheat, with a row spacing of 0.25 m. The distance between each plot was 0.5 m. The image data was collected using the DJI Phantom4 drone (SZ DJI Technology Co., Shenzhen, China) by carrying an RGB camera. The camera has a resolution of 4000 × 3000 pixels. The flight height was set to 20 m to balance the acquisition accuracy and efficiency. Concurrently, a cross-flight configuration was established, encompassing flight paths oriented along an east-west axis and a north-south axis, while the forward and side overlaps were both set as 80%, and the camera angle during the flight was set to 80°. Eleven ground control points (GCPs) were established for image quality assessment in the field.
If you find this work useful for your research, please consider citing our paper:
@article{zhang2025osnet,
title={OSNet: an oriented instance segmentation network of breeding plot extraction from UAV RGB imagery},
author={Zhang, Ruinan and Zhang, Yuanhao and Jin, Shichao and Zang, Jingrong and Zhao, Ruofan and Yao, Jiaqi and Li, Shaochen and Li, Qing and Su, Yanjun and Wu, Jin and others},
journal={Computers and Electronics in Agriculture},
volume={236},
pages={110436},
year={2025},
publisher={Elsevier}
}