FRACTAL

FRACTAL数据集是由法国国家地理和林业信息研究所创建的超大规模航空激光雷达数据集，旨在支持大规模土地监测的3D深度学习方法的发展。该数据集包含100,000个密集点云，覆盖面积达250平方公里，每个点云大小为50x50米，平均包含20k至40k个标记点，点密度为37点/平方米。数据集通过从法国全国开放的激光雷达数据中采样，特别关注稀有类别和挑战性景观，涵盖了五个法国地区的多样景观。FRACTAL数据集的应用领域包括生态监测、风险管理等公共行动，以及详细的三维地图制作，有助于现有地理数据库的巩固。

The FRACTAL dataset is a large-scale aerial LiDAR dataset created by the French National Institute of Geographic and Forestry Information, designed to advance 3D deep learning methods for large-scale land monitoring. It comprises 100,000 dense point clouds covering a total area of 250 square kilometers, with each individual point cloud spanning a 50×50 meter area and containing an average of 20,000 to 40,000 labeled points, with a point density of 37 points per square meter. Compiled by sampling from France's nationwide open LiDAR datasets, the dataset places special emphasis on rare categories and challenging landscapes, and covers diverse landscapes across five French regions. Its application areas include public initiatives such as ecological monitoring and risk management, as well as detailed 3D map production, and it helps to consolidate existing geographic databases.