AI-Augmented Context-Aware Generative Pipelines for 3D Content

This work presents a system for transforming scenes composed of basic 3D primitives into high-fidelity 3D environments using advanced representations such as Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting. The proposed approach leverages existing image stylization models and image-to-3D generative techniques to construct a pipeline that iteratively stylizes individual objects and integrates them into complex scenes. The effectiveness of the system is demonstrated through the progressive insertion of generated objects into a base scene, followed by an analysis of current limitations and potential directions for future development.