URSO (Unreal Rendered Spacecraft On-Orbit) is a simulator built on Unreal Engine 4 to render photorealistic images of spacecrafts orbiting the earth, which can be used to learn and evaluate spacecraft pose estimation, tracking and detection algorithms.
This work is part of: P. Proença and Y. Gao, Deep Learning for Spacecraft Pose Estimation from Photorealistic Rendering, ICRA, 2020 (accepted) PDF Dataset Code
Context
Spacecraft pose estimation is an important task to on-orbit proximity maneuvers in rendezvous/docking and servicing operations, but also future space debris removal missions. While this has been mostly addressed by using classic computer vision techniques based on edge-models and hand-engineered features, these methods are not robust to the challenging lighting conditions of space and the earth background. This was recently acknowledged by ESA and Stanford SLAB who joined forces to open a competition to solve this problem using deep learning.
Results
The video below gives a glimpse of how deep learning models trained on URSO datasets perform on real space imagery.
For more information and results check our paper. This is ongoing work, we plan to release more data and a version of the simulator in a near future.
