Computer vision systems for mobile autonomous machines experience a wide variety of real-world conditions and inputs that can be challenging to capture accurately in training datasets. Few autonomous systems experience more challenging conditions than those in orbit. In this talk, we’ll describe how SCOUT Space has designed and trained satellite vision systems using dynamic and physically-informed synthetic image datasets. We’ll describe how we generate synthetic data for this challenging environment, and how we leverage new real-world data to improve our datasets. In particular, we’ll explain how these synthetic datasets account for and can replicate real sources of noise and error in the orbital environment, and how we supplement them with in-space data from the first Scout-Vision system, which has been in orbit since 2021.