Assessing the Predictive Value of Physics-Grounded Synthetic Data for Computer Vision in Space Environments

This poster studies how physics-grounded synthetic data can support computer vision models for space-domain perception, where collecting labeled real imagery is difficult and operational conditions are highly constrained.

The project evaluates whether synthetic training environments can predict downstream model behavior in space-relevant visual settings, with emphasis on dataset design, domain shift, and practical value for spacecraft inspection and autonomy workflows.

• Related Research
• Contact the Lab