SemanticKITTI
收藏资源简介:
SemanticKITTI是由德国波恩大学创建的大型数据集,专注于激光雷达序列的语义场景理解。该数据集包含超过43,000次扫描,提供了28个类别的点级标注,适用于多种任务,如激光基语义分割和语义场景完成。数据集基于KITTI视觉里程计基准,覆盖了汽车激光雷达的完整360度视场,为自动驾驶等应用提供了精细的表面和物体理解。此外,数据集还提供了基准任务和基线实验,展示了现有方法在处理这些任务时的不足,为开发更先进的方法提供了丰富的数据资源。
SemanticKITTI is a large-scale dataset created by the University of Bonn, Germany, focusing on semantic scene understanding for LiDAR sequences. It contains over 43,000 scans and provides point-level annotations for 28 categories, supporting multiple tasks such as LiDAR-based semantic segmentation and semantic scene completion. Built upon the KITTI Visual Odometry Benchmark, the dataset covers the full 360-degree field of view of automotive LiDARs, enabling fine-grained surface and object understanding for applications like autonomous driving. Additionally, the dataset provides benchmark tasks and baseline experiments, which demonstrate the shortcomings of existing methods when handling these tasks, serving as a rich data resource for developing more advanced approaches.




