Abstract
In this system, we discuss methods to stylize a scene of 3D primitive objects into a higher fidelity 3D scene using novel 3D representations like NeRFs and 3D Gaussian Splatting. Our approach leverages existing image stylization systems and image-to-3D generative models to create a pipeline that iteratively stylizes and composites 3D objects into scenes. We show our results on adding generated objects into a scene and discuss limitations.