HuashuTrees: A terrestrial laser scanning point cloud dataset with detailed annotations for leaf-on and leaf-off seasons across multiple tree species

HuashuTrees is a public multi-species, bi-seasonal TLS point cloud dataset collected from deciduous broadleaf plantation forests in Nanjing, China. The leaf-off acquisition is in January 2025 and leaf-on acquisition is in July 2025. A total of seven plots (each representing a species) were scanned. The Reigl VZ400-i scanner was used at 3*3 grid pattern (9 scan positions) in each plot. This lead to highly dense point clouds. All point clouds underwent multi-station registration, denoising, individual-tree segmentation, leaf–wood separation, and manual correction. Other than plot-level point clouds with individual tree instance segmentation labels, we also provided individual tree point with leaf-wood annotation. The final tree-level dataset comprises 318 single-tree point clouds from 159 individual trees across seven deciduous broadleaf species (159 leaf-on and 159 leaf-off). All point clouds are stored in LAS/LAZ format with rich per-point attributes. Technical validation of diameter at breast height (DBH) demonstrates that point-cloud-derived estimates agree well with field measurements (RMSE =0.37 cm, R² = 0.976), with no systematic bias across seasons. HuashuTrees can be used for developing and evaluating algorithms in leaf-wood separation, individual-tree segmentation, species classification, forest structure inversion, 3D tree reconstruction and multi-temporal monitoring, serving as a key resource for forest ecology, remote sensing, and computer vision research. for more details, please visit https://github.com/qingfengxitu/HuashuTrees