Shutter Interaction Dataset Version 1.0 - Zipped

<h1 id="shutter-interaction-dataset">Shutter Interaction Dataset</h1>
<p>The Shutter Interaction Dataset is a dataset that was originally developed for predicting human intent to interact with a robot in public, but could be used for a variety of applications. The data is a rich observation space that primarily contains human and robot pose .</p>
<p>The Shutter Interaction Dataset currently consists of two smaller datasets collected in the same manner in two different indoor locations at Yale University. These two locations, called "Lobby 1" and "Lobby 2".</p>
<p>Refer to the <a href="https://shutter.interactive-machines.com/dataset/">Shutter Interaction Dataset website</a> for details on the robot, the nature of interaction, and dataset metadata.</p>
<h2 id="what-s-included">What's Included</h2>
<p>Currently, we have released processed csv files that contain human pose, robot pose, and a few features on the relative pose between the human and the robot. The data in these csv files are temporally aligned and downsampled to 5Hz. Both interaction with the robot and future interaction/interaction intent are labeled.</p>
<p>Each csv file corresponds to a time when at least one person is within view of the robot. When nobody was visible by the robot for more than 0.5 seconds, the example was terminated.</p>
<p>A diagram of each lobby with the approximate deployment location of the robot are included.</p>
<h2 id="feature-list">Feature list</h2>
<p>We use Kinect cameras to record the human poses. Many features are repeated for each joint - for conciseness, we refer to such features as <code><kinect_joint>_<feature_name></code>. This indicates that the feature <code><feature_name></code> exists for all Kinect joints. A list of Kinet joints is below:</p>
<p>| :---: | :---: |
| head |</p>
<table>
<thead>
<tr>
<th style="text-align:left">Feature Name</th>
<th style="text-align:left">Description</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:left">Python Hat</td>
</tr>
</tbody>
</table>
<h3 id="future-releases">Future releases</h3>
<p>We hope to release both the unprocessed data and processing code in early 2025 to enable the use of this data for additional research. This additional data will include information about both the way people approach and move around the robot and the interactions with the robot.</p>
<p>We also hope to continue adding to this dataset as we and others use Shutter as a research platform. As more data is collected that can be publicly shared, we plan to continue adding to this dataset.</p>
<h2 id="administrivia">Administrivia</h2>
<p>If you use these labels, please cite the original paper using this reference:</p>
<h5 id="thompson-sydney-alexander-lew-yifan-li-elizabeth-stanish-alex-huang-rohan-phanse-and-marynel-v-zquez-predicting-human-intent-to-interact-with-a-public-robot-the-people-approaching-robots-database-par-d-in-proceedings-of-the-26th-international-conference-on-multimodal-interaction-pp-536-545-2024-">Thompson, Sydney, Alexander Lew, Yifan Li, Elizabeth Stanish, Alex Huang, Rohan Phanse, and Marynel Vázquez. "Predicting Human Intent to Interact with a Public Robot: The People Approaching Robots Database (PAR-D)." In Proceedings of the 26th International Conference on Multimodal Interaction, pp. 536-545. 2024.</h5>