URSO: Unreal Rendered Spacecrafts On-Orbit Datasets

<p>Synthetic image datasets of Soyuz and Dragon spacecraft&nbsp;models orbiting the Earth, rendered using our simulator, to learn and evaluate pose estimation algorithms. For more information, please refer to our preprint:&nbsp;<a href="https://arxiv.org/abs/1907.04298">https://arxiv.org/abs/1907.04298</a></p> <p>Each dataset contains:</p> <ul> <li>5000 RGB images with a resolution of 1280x960 px</li> <li>Train/val/test set splits&nbsp;as a list: <strong>&lt;set&gt;_images.csv</strong></li> <li>Pose ground truth as: <strong>&lt;set&gt;_poses_gt.csv</strong>&nbsp;with <ul> <li>Row format:<strong> &lt;x y z qx qy qz qw&gt;</strong></li> </ul> </li> </ul> <p>The camera calibration matrix is:</p> <p><span class="math-tex">\(K = \begin{bmatrix} 640.00 &amp; 0 &amp; 640 \\ 0 &amp; 640.46 &amp; 480\\ 0 &amp; 0 &amp; 1 \end{bmatrix}\)</span></p> <p>And the horizontal and vertical field-of-view are respectively 90.0&deg; and 73.7&ordm;.</p> <p>&nbsp;</p> <p>&nbsp;</p>