CSG-Based ML-Supported 3D Translation of Sketches into Game Assets

The CSG-Based ML-Supported 3D Translation research tackles the long-standing challenge of transforming freehand 2D sketches into detailed 3D models with minimal user effort. By combining the geometric precision of Constructive Solid Geometry (CSG) with advanced machine learning techniques, this project empowers designers, architects, and game developers to bridge the gap between conceptual sketches and tangible 3D assets.

Key innovations include leveraging CSG to extrude simple geometric shapes into complex structures and integrating machine learning for precise sketch recognition. The project’s web-based interface allows users to draw freely and watch their sketches come to life in real-time 3D, revolutionizing workflows that traditionally required extensive expertise and time.

With applications in architecture, game design, and prototyping, this research streamlines creative processes, enabling faster iterations and broadening access to high-quality 3D modeling tools. The system has already demonstrated a 40% improvement in modeling automation and a 20% reduction in texture rendering time, making it a practical solution for real-world creative challenges. This work not only advances the state of sketch-to-3D translation but also opens new possibilities for intuitive and inclusive design practices.

Recommended citation: Chen, Y., Liu, Y., & Kayar-Ceylan, G. (2025). "CSG-based ML-supported 3D translation of sketches into game assets for game designers." Visual Computer. https://doi.org/10.1007/s00371-024-03758-9