Procedural Modeling of Buildings

1. Paper Title, Authors, and Affiliations

Title: Procedural Modeling of Buildings

Authors: Pascal Müller, Peter Wonka, Simon Haegler, Andreas Ulmer, Luc Van Gool

Affiliations:

• Pascal Müller, Simon Haegler, Andreas Ulmer, Luc Van Gool – ETH Zürich
• Peter Wonka – Arizona State University / ETH Zürich / K.U. Leuven

2. Main Contribution

This paper presents CGA Shape, a new way to generate procedural building models using shape grammar. It improves on past methods by ensuring architectural consistency while allowing for detailed urban environments. The key contributions include:

• A rule-based system for procedural modeling, allowing the definition of hierarchical rules that control the generation of building components.
• Context-sensitive shape grammar, enabling adaptive modifications based on spatial relationships and constraints (e.g., ensuring windows and doors align with floors and walls properly).
• New subdivision and repeat operators, which improve facade detailing and allow for realistic variation in architectural elements.
• Occlusion-aware modeling, which prevents elements from overlapping incorrectly and ensures plausible structural configurations.
• Snapping and alignment mechanisms, allowing buildings to conform to both global urban layouts and fine-grained local adjustments.
• Scalability for large-scale urban environments, demonstrated through examples of real-world city modeling, including historical reconstructions and modern cityscapes.

3. Outline of the Major Topics

3.1 Introduction and Motivation

• Problems in procedural city modeling.
• Weaknesses of past approaches (e.g., L-systems, simple splitting methods).

3.2 CGA Shape: A New Approach

• Using shape rules to define buildings.
• How context-aware constraints improve realism.

3.3 Core Techniques

• Scope Rules: Define building parts and transformations.
• Splitting and Repeating: Generate facades, floors, and repeating patterns.
• Occlusion Checks: Ensure elements do not intersect incorrectly.
• Snapping: Align elements to maintain structure.

3.4 Implementation and Examples

• Applied to historical and modern cityscapes (e.g., Pompeii, Beverly Hills).
• Discussion on performance and scalability.

3.5 Conclusion and Future Work

• Plans to improve real-time rendering.
• Potential extensions for interior modeling.

4. One Thing I Liked

One of the most useful aspects of the paper is the integration of procedural modeling with geometric constraints. Unlike older methods that generate buildings without structural awareness, CGA Shape ensures elements like doors, windows, and floors align properly. This makes it particularly useful for large-scale city generation where both efficiency and realism are needed.

5. What I Did Not Like

While the paper presents a strong procedural approach, it mainly focuses on exteriors and does not discuss interior layouts in depth. For real-world applications, integrating interior spaces would make the system much more useful for architects and urban planners. Additionally, performance trade-offs for extremely detailed models could have been explored more.

6. Questions for the Authors

1. How well does CGA Shape handle non-standard building shapes? Many modern buildings have curved or irregular designs—does the system require special rules for these cases?
2. Could CGA Shape be extended to model building interiors? While great for exteriors, could similar rules be applied for floorplans and room layouts?
3. What are the memory and computational limits for large-scale city models? The paper mentions billion-polygon cities, but what are the practical constraints for real-time applications?
4. Could CGA Shape work with AI-based techniques? With recent advances in AI-generated content, do you see a way to integrate machine learning to optimize or automate rule generation?