Aachen Day-Night, RobotCar Seasons, CMU Seasons
收藏资源简介:
本研究引入了三个新的户外基准数据集,包括Aachen Day-Night、RobotCar Seasons和CMU Seasons,专门用于评估在变化条件下如光照(日/夜)、天气(晴/雨/雪)和季节(夏/冬)的6自由度视觉定位。这些数据集覆盖了多种场景,如行人定位和车辆定位,以及单图像和多图像序列的定位。通过这些数据集,研究者能够首次分析这些变化条件对6自由度相机姿势估计精度的影响。数据集的创建依赖于人工标注的图像匹配和验证的地面实况姿势,确保了数据集的质量和准确性。这些数据集的应用领域广泛,旨在解决自动驾驶车辆导航和增强现实应用中的视觉定位问题。
This study introduces three novel outdoor benchmark datasets, namely Aachen Day-Night, RobotCar Seasons, and CMU Seasons, specifically designed for evaluating 6-degree-of-freedom (6-DoF) visual localization under varying conditions including illumination (day/night), weather (sunny/rainy/snowy), and seasons (summer/winter). These datasets cover diverse scenarios such as pedestrian localization and vehicle localization, as well as localization tasks with single images and multi-image sequences. With these datasets, researchers can, for the first time, analyze the impact of these varying conditions on the accuracy of 6-degree-of-freedom camera pose estimation. The creation of these datasets relies on manually annotated image matches and verified ground-truth poses, which ensures the quality and accuracy of the datasets. These datasets have broad application prospects, aiming to address visual localization challenges in autonomous vehicle navigation and augmented reality applications.




