MUSES

该数据集是一个专为在具有挑战性的环境条件下，自动驾驶中的密集语义感知而设计的多模态基准。它包含了高质量的2D全景标注，覆盖了19个语义类别，规模达到了2500个样本。该数据集的任务包括语义分割、全景分割以及不确定性感知全景分割。

This dataset is a multimodal benchmark specifically designed for dense semantic perception in autonomous driving under challenging environmental conditions. It contains high-quality 2D panoramic annotations covering 19 semantic categories, with a total of 2500 samples. The supported tasks include semantic segmentation, panoptic segmentation, and uncertainty-aware panoptic segmentation.