MCD (Multi-Campus Dataset)

该数据集名为MCD，包含了广泛的感知模式、高精度的地面真实数据，以及跨越三个欧亚大学校园的多样化挑战性环境。在这59,000个稀疏NRE激光雷达扫描中，对29个类别进行了语义标注。数据集不仅包含白天和夜晚的序列，允许在不同光照条件下评估算法，而且还基于激光雷达-惯性数据优化提供了连续时间的地面真实数据。该数据集包含18个序列，环境变化高度多样。其任务是对不同条件下SLAM方法的语义分割和评估。The dataset is named MCD. It encompasses a wide range of perceptual modalities, high-precision ground truth data, and diverse challenging environments spanning three Eurasian university campuses. Of its 59,000 sparse NRE LiDAR scans, 29 categories are annotated with semantic labels. The dataset not only includes daytime and nighttime sequences, enabling the evaluation of algorithms under varying lighting conditions, but also provides continuous-time ground truth data optimized based on LiDAR-inertial data. It consists of 18 sequences with highly diverse environmental variations. The supported tasks cover semantic segmentation and performance evaluation of SLAM methods under different conditions.

搜集汇总

背景与挑战

背景概述

MCD (Multi-Campus Dataset) 是一个用于SLAM方法语义分割和评估的多模态数据集，涵盖三个欧亚大学校园的多样化环境，包含59,000个稀疏NRE激光雷达扫描和29个类别的语义标注。数据集提供白天和夜晚序列，支持在不同光照条件下进行算法评估，并基于激光雷达-惯性数据优化提供连续时间的地面真实数据，共包含18个高度变化的序列。