Panoptic nuScenes
收藏资源简介:
Panoptic nuScenes是由慕尼黑工业大学开发的一个大规模激光雷达全景分割和跟踪基准数据集。该数据集扩展自流行的nuScenes数据集,包含了1000个场景,覆盖新加坡和波士顿的多个地点,具有高度的地理多样性。数据集包含11亿个标注的激光雷达点,涵盖32个语义类别,用于支持自动驾驶中的全景场景理解任务。Panoptic nuScenes不仅包含静态场景的分割,还特别关注动态对象的跟踪,为全景跟踪任务提供了丰富的实例和时间一致的标注。此数据集的应用领域主要集中在自动驾驶技术,旨在解决复杂城市环境中动态对象的准确识别和跟踪问题。
Panoptic nuScenes is a large-scale LiDAR panoptic segmentation and tracking benchmark dataset developed by the Technical University of Munich. It is extended from the popular nuScenes dataset, containing 1000 scenes collected across multiple locations in Singapore and Boston, with high geographic diversity. The dataset includes 1.1 billion annotated LiDAR points spanning 32 semantic categories, supporting panoptic scene understanding tasks in autonomous driving. Panoptic nuScenes not only covers segmentation of static scenes but also places special emphasis on tracking dynamic objects, providing rich instance-level and temporally consistent annotations for panoptic tracking tasks. Its application fields are mainly focused on autonomous driving technology, aiming to solve the problems of accurate recognition and tracking of dynamic objects in complex urban environments.

- 1Panoptic nuScenes: A Large-Scale Benchmark for LiDAR Panoptic Segmentation and Tracking慕尼黑工业大学 · 2021年



