Livox Simu-dataset

Livox仿真数据集基于自动驾驶仿真测试平台生成的点云数据和相应的注释，支持3D对象检测和语义分割任务。传感器配备了5个豪杰地平线激光雷达和1个Tailan Tele-15超远程激光雷达。整个数据集包含14,445帧的360 ° Lidar点云数据，6种类型的3D边界框注释和14种类型的目标。类别的点云语义注释。这次开放下载的数据集主要是城市宽道路场景，包括双向12车道和双向8车道。相应地，模拟场景还包括多种车辆和行人模型，以及更接近真实场景的交通流模拟。此外，丰富的交通信号灯、交通标志、隔板 (包括隔离栏杆、绿化带、隔离墩等) 、树木建筑等使整个模拟场景更接近实际驾驶情况。

The Livox Simulation Dataset comprises point cloud data and corresponding annotations generated via an autonomous driving simulation test platform, supporting 3D object detection and semantic segmentation tasks. The sensor setup includes 5 Horizon Livox lidars and 1 Tailan Tele-15 ultra-long-range lidar. The full dataset contains 14,445 frames of 360° LiDAR point cloud data, with 6 types of 3D bounding box annotations and 14 object categories accompanied by point cloud semantic annotations. The currently publicly released dataset primarily covers urban wide-road scenarios, including 12-lane and 8-lane bidirectional roadways. Correspondingly, the simulation scenarios incorporate diverse vehicle and pedestrian models, alongside traffic flow simulations that closely mimic real-world conditions. Furthermore, abundant elements such as traffic lights, traffic signs, barriers (including isolation railings, green belts, isolation piers, etc.), trees, and buildings render the entire simulation scenario highly consistent with actual driving environments.

搜集汇总

背景与挑战

背景概述

Livox Simu-dataset是一个公开的自动驾驶领域数据集，由OpenDataLab提供，大小为30.9GB，采用GPL-3.0许可证。但数据集详情页未提供具体介绍，因此无法总结其内容、特点或用途。