LaRS
收藏资源简介:
LaRS数据集由卢布尔雅那大学的ViCoS实验室创建,旨在解决海上障碍物检测领域的数据多样性和复杂性问题。该数据集包含超过4000个关键帧,每个关键帧都配有全景分割标注,涵盖8个物体类别和3个背景类别,以及19个描述环境类型、光照条件、反射水平等的全局场景属性。此外,每个关键帧还附带了前9帧,以支持利用时间纹理的方法开发。LaRS数据集的应用领域包括自主无人水面舰艇(USV)的障碍物检测,旨在提高检测的准确性和鲁棒性,特别是在复杂和多变的海上环境中。
The LaRS dataset was developed by the ViCoS Laboratory at the University of Ljubljana, with the aim of addressing the challenges of data diversity and complexity in maritime obstacle detection. This dataset includes over 4,000 key frames, each annotated with panoptic segmentation labels, covering 8 object categories, 3 background categories, and 19 global scene attributes that describe environmental types, lighting conditions, reflection levels and other relevant characteristics. Furthermore, each key frame is accompanied by its preceding 9 consecutive frames to support the development of methods leveraging temporal texture information. The LaRS dataset is applied to obstacle detection for autonomous unmanned surface vehicles (USV), with the objective of improving detection accuracy and robustness, especially in complex and dynamic maritime environments.

- 1LaRS: A Diverse Panoptic Maritime Obstacle Detection Dataset and Benchmark卢布尔雅那大学 · 2023年



