Next-Generation Spacecraft Pose Estimation Dataset (SPEED+)

SPEED+ is an advanced dataset for vision-based spacecraft pose estimation with specific emphasis on evaluating the robustness of Machine Learning (ML) models across the domain gap. It includes images of the Tango spacecraft from the PRISMA mission and features three domains from two sources:
1. Synthetic Domain: Created with Stanford's OpenGL-based Optical Stimulator, it consists of 59,960 labeled synthetic images split 80:20 for training/validation.
2. Hardware-In-the-Loop (HIL) Domains: 1. Generated at SLAB's Testbed for Rendezvous and Optical Navigation (TRON) facility, using lightboxes and sunlamp for realistic illumination. These domains are accompanied with high-accuracy pose labels recovered from the calibrated TRON facility. In practice, they are reserved for testing, reflecting real-world scenarios without pre-labeled target space images.

SPEED+ is publicly available and part of the second international Satellite Pose Estimation Competition (SPEC2021), co-hosted by SLAB and the European Space Agency's Advanced Concepts Team. The applications of autonomous vision-based spaceborne navigation using the SPEED+ dataset are extensive and crucial for various space missions.

Figure

Figure description - Example images from different domains of SPEED+ DOI: 10.1109/AERO53065.2022.9843439