Hedgementation
收藏资源简介:
Hedgementation是由不列颠哥伦比亚大学等机构联合构建的遥感基准数据集,专注于法国全国范围的树篱分割任务。该数据集整合了Sentinel-2卫星影像、Alpha Earth Foundations嵌入特征以及法国国家地理研究所的BD Haie树栅标签,共包含2,995个1.28km×1.28km的影像块,总计49,070,080像素,其中树栅像素占比4.95%。数据创建过程涉及多源数据融合、空间栅格化处理及气候分区标注,通过缓冲区分割确保训练与测试样本的空间独立性。该数据集旨在评估机器学习模型在跨空间距离和农业气候区的泛化能力,为解决大尺度生态环境监测和农业景观管理提供关键基准。
Hedgementation is a remote sensing benchmark dataset jointly developed by institutions including the University of British Columbia, focusing on nationwide hedge segmentation tasks in France. This dataset integrates Sentinel-2 satellite imagery, embedded features from Alpha Earth Foundations, and BD Haie hedge labels from the French National Geographic Institute. It comprises 2,995 image tiles each sized 1.28 km × 1.28 km, with a total of 49,070,080 pixels, among which hedge pixels account for 4.95%. The dataset creation process involves multi-source data fusion, spatial rasterization processing, and climate zone annotation, and buffer segmentation is used to ensure the spatial independence of training and test samples. This dataset aims to evaluate the generalization ability of machine learning models across spatial distances and agroclimatic zones, providing a critical benchmark for addressing large-scale ecological environment monitoring and agricultural landscape management.
数据集概述:Hedgementation Benchmark
名称:Hedgementation Benchmark
类型:遥感图像分割基准(Hedgerow Segmentation,树篱分割)
论文:对应发表在 4th ICLR Workshop on Machine Learning for Remote Sensing (Main Track) 的论文《Hedgementation = Hedgerow Segmentation: A Remote Sensing Benchmark》,论文地址为:http://arxiv.org/abs/2606.23615
仓库地址:https://github.com/hedgementation/hedgementation
数据集构成与使用流程
该仓库包含构建和使用 Hedgementation 基准所需的完整代码,分为三个子目录:
-
Dataset Construction(数据集构建)
- 包含从头开始完整创建数据集的所有代码。
- 目录下 README.md 中有详细的构建与导出说明。
-
Model Training(模型训练)
- 提供在 Hedgementation 数据集上训练模型的完整框架。
- 目录下 README.md 中有快速入门指南。
-
Hedgementation Utils(通用工具)
- 存储两个子项目共享的实用功能代码。
使用方法
- 首先进入
dataset_construction目录,按照说明构建、导出并下载数据集。 - 然后进入
model_training目录,按照指引在本地数据上训练模型。
引用方式
若使用该基准,请引用以下 BibTeX:
bibtex @inproceedings{senyard2026hedgementation, title = {Hedgementation = Hedgerow Segmentation: A Remote Sensing Benchmark}, author = {Senyard, Nathan and Hamdani, Salem and Zhang, Astrid and Shelhamer, Evan and L{e}cuyer, Mathias and Gantois, Jos{e}phine}, booktitle = {4th ICLR Workshop on Machine Learning for Remote Sensing (Main Track)}, year = {2026}, url = {https://openreview.net/forum?id=mOMTBBgq5n} }




