Academic Research
Development & Research
Exploring the bleeding edge of 3D Gaussian Splatting and Generative AI. Here I present the findings from my academic work at Hochschule Düsseldorf, focusing on how AI transforms the way we create and edit 3D environments.
AI-Driven Generation and Modification of 3D Scenes from Single Images
My recent academic paper explores the paradigm shift in computer graphics: Moving from manual 3D modeling to AI-driven generation and editing using 3D Gaussian Splatting (3DGS). The research evaluates current State-of-the-Art frameworks and their applicability for broadcast and VR.
1. Single-Image to 3D
Analyzing the transformation of flat 2D images into navigable 3D spaces using monocular depth estimation, Outpainting, and Feed-Forward 3DGS architectures like Splatter Image.
2. AI-Driven Scene Editing
Evaluating prompt-based object insertion and removal. Highlighting how frameworks like GaussianEditor use Semantic Tracing to edit specific 3D regions without disrupting the global scene consistency.
3. Geometry & Texture
Exploring the disentanglement of appearance and structure. Using Texture-GSand SuGaR to extract editable polygon meshes from Gaussians, enabling true integration into classical rendering workflows.
Bachelor Thesis (Upcoming)
Building upon the theoretical foundation of the "Scientific Specialization", my upcoming Bachelor Thesis will conduct a comprehensive empirical evaluation of these AI frameworks.