WHU-STree_多模态行道树基准数据集

WHU-STree是一个开创性的多模态数据集，旨在弥补城市街树清查研究中的关键空白。通过整合同步点云和在两个地理上截然不同的城市收集的高分辨率图像，WHU-STree为推动自动化城市林业管理提供了全面的资源。该数据集包含 21,007 个带注释的树木实例，并带有 50 种树种和形态参数的物种标签，支持单棵树分段、物种分类和三维形态分析等任务。其跨城采集策略结合丰富的注释和多模态集成，为评估算法泛化和鲁棒性在现实城市环境中树立了新标杆。

WHU-STree is a pioneering multimodal dataset aimed at bridging critical gaps in urban street tree inventory research. By integrating synchronized point clouds and high-resolution imagery collected from two geographically distinct cities, WHU-STree provides a comprehensive resource to advance automated urban forestry management. This dataset contains 21,007 annotated tree instances, with species labels for 50 tree species and morphological parameters, supporting tasks including individual tree segmentation, species classification, and 3D morphological analysis. Its cross-city collection strategy, combined with rich annotations and multimodal integration, sets a new benchmark for evaluating algorithm generalization and robustness in real-world urban environments.

搜集汇总

背景与挑战

背景概述

WHU-STree是一个多模态行道树基准数据集，整合了来自南京和沈阳的点云与高分辨率图像，包含21,007个标注树实例、50个树种和形态参数，用于支持个体树分割、物种分类和3D形态分析等任务。该数据集通过跨城市收集策略和丰富标注，为评估算法在真实城市环境中的泛化性和鲁棒性提供了新基准。