Human Morning Routine Dataset

This dataset has been created with the aim of capturing a typical morning routine of a single person in an apartment. The focus for this dataset was to capture several activities of daily living performed by the same person at different days during an extended period of time. It includes diary-like notes of the male participant in his rental apartment including the times and the order in which particular tasks were executed during 14 working days. Furthermore, the dataset features motion tracking data of a sensor equipped kitchen environment in which the participant was asked to reenact his morning routine. The same kitchen environment has also been rebuilt in the MORSE Simulator and sensor-data of the reenactment in the simulated environment is included as well. The experiments were manually labelled to provide ground truth data for the evaluation of potential experiments.

Furthermore, motion tracking data was generated by reenacting the morning routine activities that took place in the kitchen. For this dataset, the same scenario as in the TUM kitchen dataset was used, but we specifically focused on a low cost sensor setting using only two Kinect sensors which cost below 250 USD in contrast to relying on complex motion tracking systems that are very costly. We decided to go for such a sensor setting since we think that a commercial service robot that will be deployed on the consumer market will use rather low cost components in order to be affordable for a wide range of people. Thus our dataset reflects a realistic impression about how such a robot would perceive routine activities of its user when it is deployed in a human apartment.

A detailed description of the dataset can be found in the README.pdf

The full dataset can be downloaded in one .tar.gz-file from the following link: Download

In case you use this dataset for a publication, please provide a citation to the following paper in which the dataset is described:

A Human Morning Routine Dataset (Michael Karg, Alexandra Kirsch), In Proceedings of the Ninth International Conference on Autonomous Agents and Multiagent Systems (AAMAS), Extended Abstract, 2014. [pdf]

Related Publications

Low Cost Activity Recognition Using Depth Cameras and Context Dependent Spatial Regions (Michael Karg, Alexandra Kirsch), In Workshop on Autonomous Robots and Multirobot Systems (ARMS), Proceedings of the Ninth International Conference on Autonomous Agents and Multiagent Systems (AAMAS), 2014. [pdf]

Low Cost Activity Recognition Using Depth Cameras and Context Dependent Spatial Regions (Michael Karg, Alexandra Kirsch), In Proceedings of the Ninth International Conference on Autonomous Agents and Multiagent Systems (AAMAS), Extended Abstract, 2014. [pdf]

Simultaneous Plan Recognition and Monitoring (SPRAM) for Robot Assistants (Michael Karg, Alexandra Kirsch), In Proceedings of Human Robot Collaboration Workshop at Robotics Science and Systems Conference (RSS) 2013, 2013. [pdf]

An Expectations Framework for Domestic Robot Assistants (Michael Karg, Alexandra Kirsch), In Conference on Advances in Cognitive Systems, 2013. [pdf]

Acquisition and Use of Transferable, Spatio-Temporal Plan Representations for Human-Robot Interaction (Michael Karg, Alexandra Kirsch), In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2012. [pdf]