LeafMachine2 Data: Juglandaceae Leaf Outlines and ECT

Juglandaceae - 517,187 leaves - 67.4 GB: This repository contains HDF5 (.h5) files with leaf morphological data for species in the Juglandaceae family. Each file represents a single leaf specimen and contains leaf shape outlines, the 128x128 ECT matrix, and associated metadata. File Contents This dataset contains 517,187 leaves from taxa in the family Juglandaceae. Some leaves may be partial, predated, broken, or incomplete. Each .h5 file contains the following datasets: ECT_matrices/ - Euler Characteristic Transform (ECT) matrices capturing topological features of leaf shapes shapes/shape_0 - Coordinate array (x,y) defining the leaf outline boundary component_names - Original filename identifier (without extension) group_labels - Taxonomic classification dictionary containing: family: Taxonomic family name genus: Genus classification genus_species: Binomial species name fullname: Complete taxonomic identifier Data Format Files are organized with standardized naming: [Herbarium]_[ID]_[Family]_[Genus]_[Species]__[LeafID].h5 Shape coordinates are normalized to a unit circle centered at the origin (-0.5 to 0.5 range), vertically oriented. Uncompressed Size 67.4 GB Usage Code for reading, processing, and analyzing these files is available at: https://github.com/Gene-Weaver/LM2-Data-Tools The repository includes functions for extracting data and generating visualizations. Citation Please cite the LeafMachine2 paper and this dataset. Weaver, W. N., & Smith, S. A. (2023). From leaves to labels: Building modular machine learning networks for rapid herbarium specimen analysis with LeafMachine2. Applications in Plant Sciences, 11(5), e11548. https://doi.org/10.1002/aps3.11548胡桃科（Juglandaceae）数据集——含517,187份叶片样本，总数据量67.4 GB：本存储库包含用于存储胡桃科物种叶片形态学数据的HDF5（.h5）格式文件，每份文件对应单份叶片标本，内含叶片轮廓、128×128欧拉特征变换（Euler Characteristic Transform, ECT）矩阵及相关元数据。 ## 文件内容 本数据集涵盖胡桃科类群的517,187份叶片样本，部分叶片可能存在部分缺失、遭啃食、破损或不完整的情况。每份.h5文件均包含以下数据集： - `ECT_matrices/`：存储叶片形态拓扑特征的欧拉特征变换（ECT）矩阵 - `shapes/shape_0`：定义叶片轮廓边界的(x,y)坐标数组 - `component_names`：原始文件名标识符（不含扩展名） - `group_labels`：分类学标注字典，包含以下字段： - `family`：分类学科名 - `genus`：属级分类信息 - `genus_species`：双名法物种名称 - `fullname`：完整分类学标识符 ## 数据格式 文件采用标准化命名规则：`[标本馆]_[编号]_[科名]_[属名]_[物种名]__[叶片ID].h5`。 叶片轮廓坐标已归一化至以原点为中心的单位圆范围内（取值区间为-0.5至0.5），且为垂直朝向。 ## 未压缩数据总量 67.4 GB ## 使用说明 用于读取、处理与分析此类文件的代码可通过以下链接获取：https://github.com/Gene-Weaver/LM2-Data-Tools。该存储库内置数据提取与可视化生成相关功能函数。 ## 引用要求 请引用《LeafMachine2》相关论文及本数据集： Weaver, W. N., & Smith, S. A. (2023). From leaves to labels: Building modular machine learning networks for rapid herbarium specimen analysis with LeafMachine2. Applications in Plant Sciences, 11(5), e11548. https://doi.org/10.1002/aps3.11548