University of Michigan North Campus long-term vision and lidar dataset

Abstract

This paper documents a large scale, long-term autonomy dataset for robotics research collected on the University of Michigan's North Campus. The dataset consists of omnidirectional imagery, 3D lidar, planar lidar, GPS, and proprioceptive sensors for odometry collected using a Segway robot. The dataset was collected to facilitate research focusing on long-term autonomous operation in changing environments. The dataset is composed of 27 sessions spaced approximately biweekly over the course of 15 months. The sessions repeatedly explore the campus, both indoors and outdoors, on varying trajectories, and at different times of the day across all four seasons. This allows the dataset to capture many challenging elements including: moving obstacles (e.g. pedestrians, bicyclists and cars), changing lighting, varying viewpoint, seasonal and weather changes (e.g. falling leaves and snow), and long-term structural changes caused by construction projects. To further facilitate research, we also provide ground-truth pose for all sessions in a single frame of reference.

Image features for visual teach-and-repeat navigation in changing environments

“…The Stromovka dataset comprises of 1000 images captured during two 1.3 km long tele-operated runs in the Stromovka forest park in Prague during summer and winter 2011 [54]. The third and fourth datasets, called 'Michigan' and 'North Campus', were gathered around the University of Michigan North Campus during 2012 and 2013 [20]. Similarly to the datasets gathered in Prague, the Michigan set covers seasonal changes in a few locations over one year and the North Campus dataset consists of two challenging image sequences captured in winter and summer.…”

“…The team of the Michigan university carried on with their data collection efforts and made their 'North Campus Long-Term Dataset' publicly available [20]. This large-scale, long-term dataset consists of omnidirectional imagery, 3D lidar, planar lidar, and proprioceptive sensory data and ground truth poses, which makes it a very useful dataset for research regarding long-term autonomous navigation.…”

“…Moreover, we perform a comparative analysis of training schemes, leading to computationally-efficient image features that can deal with naturally-occurring environment changes. We apply our evaluation on a new dataset, which became available only recently [20]. Finally, we provide the aforementioned benchmarking framework and the GRIEF training method as a documented, open-source software package [21].…”

Will It Last? Learning Stable Features for Long-Term Visual Localization

“…We have also used the public University of Michigan NCLT datasets [1]. We have selected an area that was covered by 19 datasets in total (from (x, y) = (−301, −448) to (−183, −432) according to the ground-truth annotation), collected over 15 months.…”

“…Random selection: Landmarks are assigned a score that is obtained by random sampling from a uniform distribution in range [0,1]. Random selection provides a very good baseline to evaluate feature selection methods.…”