Accurate pose estimation of non-cooperative spacecraft is a key capability for future autonomous operations, including Active Debris Removal (ADR), On-Orbit Servicing (OOS) and inspection missions. Traditional computer vision methods often struggle in this domain due to the scarcity of real annotated data and the wide variability in spacecraft geometry, illumination and viewing conditions. This thesis presents a study and implementation for the generation and validation of synthetic datasets aimed at training deep learning models for spacecraft pose estimation. The system is based on a simulator developed in Unreal Engine, capable of rendering photorealistic images of a target spacecraft under diverse conditions such as lighting, background, attitude and camera configurations.