The MIT Stata Center Dataset

TitleThe MIT Stata Center Dataset
Publication TypeJournal Article
Year of Publication2013
AuthorsFallon MF, Johnnsson H, Kaess M, Leonard JJ
Volume32
Pagination1695-1699
Abstract

This paper presents a large scale dataset of vision (stereo and RGB-D), laser and proprioceptive data collected over an extended duration by a Willow Garage PR2 robot in the 10 story MIT Stata Center As of September 2012 the dataset comprises over 23 TB, 38 h and 42 km (the length of a marathon) The dataset is of particular interest to robotics and computer vision researchers interested in long-term autonomy It is expected to be useful in a variety of research areas - robotic mapping (long-term, visual, RGB-D or laser), change detection in indoor environments, human pattern analysis, long-term path planning For ease of use the original ROS ‘bag’ log files are provided and also a derivative version combining human readable data and imagery in standard formats Of particular importance, this dataset also includes ground-truth position estimates of the robot at every instance (to typical accuracy of 2 cm) using as-built floor-plans - which were carefully extracted using our software tools The provision of ground-truth for such a large dataset enables more meaningful comparison between algorithms than has previously been possible