Researcher · MSc Earth and Space Science · York University

Physics-Constrained Masked Diffusion for Space Situational Awareness

Simulation environments for optical space situational awareness imagery have difficulties modeling the higher-order noise within the environment. This lack of noise within imagery leads to lackluster results in the training of AI-based detectors purely on simulated data compared to training on in-situ data. The plan is to pair a physics-based simulation environment with the noise learning capabilities of diffusion models to develop synthetic imagery with better learning properties.