4Seasons
收藏资源简介:
4Seasons数据集是由慕尼黑工业大学计算机科学系创建的大型视觉SLAM和长期定位基准,专为自动驾驶在挑战性条件下的应用设计。该数据集包含超过300公里的记录,涵盖九种不同的环境,从多层停车场到城市(包括隧道)、乡村和高速公路。数据集提供了全球一致的参考姿态,精度高达厘米级,这些姿态是通过直接立体惯性里程计与RTK GNSS融合得到的。4Seasons数据集旨在评估视觉里程计、全局位置识别和基于地图的视觉定位性能,特别关注跨季节和多天气条件下的变化,以推动自动驾驶技术的研究和发展。
The 4Seasons dataset is a large-scale visual SLAM and long-term localization benchmark created by the Department of Computer Science of the Technical University of Munich, specifically designed for autonomous driving applications under challenging conditions. This dataset contains over 300 kilometers of recorded data, covering nine distinct environments ranging from multi-story parking lots to urban areas (including tunnels), rural regions, and highways. The dataset provides globally consistent reference poses with centimeter-level accuracy, which are derived from the fusion of direct stereo-inertial odometry and RTK GNSS. The 4Seasons dataset aims to evaluate the performance of visual odometry, global place recognition and map-based visual localization, with particular focus on variations across seasons and diverse weather conditions, so as to advance the research and development of autonomous driving technologies.




