Researchers have developed CityPaint, a new framework designed to generate high-fidelity, city-scale 3D textures for applications like virtual reality and urban digital twins. Unlike previous methods that struggled with the scarcity of 3D training data, CityPaint leverages abundant 2D image data and 2D diffusion models. The system integrates geometric priors with generative capabilities, using a novel camera baking strategy for iterative texture projection and a decoupled refinement mechanism to optimize terrain and building textures separately. Experiments indicate CityPaint produces globally coherent textures with fine local details, outperforming existing baselines in visual quality and consistency. AI
IMPACT This research could improve the realism and efficiency of creating 3D environments for gaming and virtual reality applications.
RANK_REASON The item describes a novel research framework and its technical details, published in a venue focused on research papers. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
- 2D diffusion models
- 2D image data
- 2D-to-3D projection
- 3D Texturing of the Water–Air Interface by Biomimetic Self-Assembly
- building textures
- camera baking
- CityPaint
- gambling
- Generative Models
- Geometric priors
- terrain textures
- Urban Digital Twins – A FIWARE-based model
- virtual reality
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
