Freiburg dataset

该数据集是从COLD数据库中选取的一个子集，包含了在弗莱堡大学建筑物内不同空间由机器人全方位拍摄的图片。这些图片是在真实操作条件下捕获的。此外，该数据集还包括通过激光传感器获取的地面真实数据，用于量化定位误差。基准训练数据集由556张图片组成，而测试数据集则包括多云测试数据集、晴朗测试数据集和夜晚测试数据集，分别包含2595张、2114张和2707张图片。该数据集涵盖了九个不同房间的图片以及多种光照条件。其任务是进行移动机器人的分层定位。

This dataset is a subset selected from the COLD database, containing omnidirectional images captured by robots in various spaces within the buildings of the University of Freiburg. These images were collected under real-world operational conditions. Furthermore, the dataset includes ground-truth data acquired via laser sensors, which is used to quantify localization errors. The baseline training dataset consists of 556 images, while the test datasets include the cloudy test dataset, sunny test dataset and nighttime test dataset, containing 2595, 2114 and 2707 images respectively. This dataset covers images from nine distinct rooms and a variety of lighting conditions. The task of this dataset is hierarchical localization for mobile robots.

搜集汇总

背景与挑战

背景概述

该数据集是Freiburg数据集，用于移动机器人室内定位研究，包含训练、验证和测试集，并应用了六种数据增强效果。数据集主要用于评估CNN模型在层次化室内定位中的性能。