ParisLuco3D

ParisLuco3D数据集是由国立巴黎高等矿业学院-PSL大学创建的高质量目标数据集，专门设计用于激光雷达感知的领域泛化评估。该数据集包含7501个扫描，覆盖了巴黎市中心的多样化场景和上下文，使用Velodyne HDL32激光雷达传感器采集。数据集不仅用于评估激光雷达语义分割、物体检测和跟踪的性能，还提供了在线基准，确保方法间的公平比较。ParisLuco3D数据集的创建旨在解决现有数据集在跨领域评估中存在的标注策略差异问题，通过提供精细的标注，避免在领域泛化方法的质量评估中做出妥协。

The ParisLuco3D dataset is a high-quality annotated dataset developed by Mines Paris – PSL University, specifically tailored for domain generalization evaluation in LiDAR perception. It comprises 7501 scans captured across diverse scenarios and contextual settings in downtown Paris, using a Velodyne HDL32 LiDAR sensor. This dataset supports the evaluation of LiDAR-based semantic segmentation, object detection and tracking performance, while also offering an online benchmark to enable fair comparisons between different research methods. The ParisLuco3D dataset was designed to resolve the issue of inconsistent annotation strategies across existing datasets used for cross-domain evaluation, by providing fine-grained annotations to avoid compromising the quality assessment of domain generalization approaches.