Annotated 3D Point Cloud Dataset of Broad-Leaf Legumes Captured by High-Throughput Phenotyping Platform.

Abstract

This data descriptor presents novel, annotated 3D point cloud plant scans generated by a high-throughput phenotyping platform (LeasyScan, ICRISAT, India). It focuses on broad-leaf legume species (mungbean, common bean, cowpea, and lima bean). The dataset, generated by PlantEye(R) F600 technology, captures multispectral 3D scans of plant canopies. It includes 223 scans, providing detailed organ-level segmentation annotations for embryonic leaves, leaves, petioles, stems, and whole plants. The dataset fills a critical gap in plant phenomics research by offering a base of annotated data to support AI model development efforts in 3D computer vision. Data preprocessing, annotation procedures, and potential applications in crop research disciplines are further discussed. The dataset, preprocessing code, annotations, and a MIAPPE-compliant data sheet are also presented via the GitHub repository for further updates and expansion.

Plant phenotyping relevance

植物フェノタイピングプラットフォームで取得した3D点群と器官レベル注釈を提供するデータセットで、再利用可能な画像解析・AI開発基盤が中心です。

Code and data availability

The paper's own annotated 3D point cloud dataset (223 scans of legumes with organ-level segmentation annotations), raw scanner data, MIAPPE metadata, and preprocessing/cuboid-generation/baseline-evaluation code are publicly deposited on Figshare and mirrored on GitHub.