Pi3DET
收藏资源简介:
Pi3DET是一个多平台3D物体检测数据集,由超过51,545个激光雷达帧和超过25万个精心标注的3D边界框组成,涵盖了车辆、四足动物和无人机等多个平台。该数据集是在多个环境中收集的,包括城市、郊区和农村地区,以及白天和夜晚的条件。Pi3DET的创建使用了一个自动标注流程,并辅以大约500小时的广泛手动精炼。该数据集旨在为非车辆平台以及跨平台的3D检测研究提供坚实的基础。
Pi3DET is a multi-platform 3D object detection dataset consisting of over 51,545 LiDAR frames and more than 250,000 meticulously annotated 3D bounding boxes, covering multiple platforms such as vehicles, quadrupeds, and unmanned aerial vehicles (drones). The dataset was collected across diverse environments including urban, suburban, and rural areas, under both daytime and nighttime conditions. The development of Pi3DET employed an automatic annotation pipeline, supplemented by extensive manual refinement totaling approximately 500 hours. This dataset aims to provide a robust foundation for 3D detection research targeting non-vehicle platforms as well as cross-platform scenarios.
- 1Perspective-Invariant 3D Object Detection中国科学院沈阳自动化研究所 · 2025年



