_edited.png)
LumiModeling
Project:
Date:
Keywords:
4D Reconstruction of Architectural Spaces Through Dynamic Lighting
2025.5
3D Reconstruction, Gaussian Splatting, Computer Vision
This thesis presents a 4D reconstruction framework that employs the Gaussian Splatting technique to visualize architectural environments under diverse lighting and environmental scenarios in real time. By extracting geometric and photometric information from multi-view images, the system generates relightable 3D scenes, enabling high-fidelity reconstructions.
Background
Figure 1: Renderings of Ronchamp © yane markulev
Lighting plays a fundamental role in shaping architectural atmosphere, influencing how spaces are perceived and experienced. However, current digital design tools often fall short in capturing the dynamic interplay between light, material, and space, especially under real-world and time-varying conditions. Right now, while the modeling + rendering pipeline can produce photorealistic results, it might still loss details of information compared with 3D scanning.
Research Idea
In contrast, 3D reconstruction and scanning techniques derive spatial and material properties directly from real images or videos. This research aims to develop a tool that captures architectural atmosphere—how spaces appear under varying lighting and environmental conditions.
Conclusion
In conclusion, I see this tool can be further applied in scenarios such as: architecture case studies on light and space, site-based architectural analysis, integrating real-world environmental conditions, 3D reconstruction in uncontrolled lighting conditions. Light embodies both quantifiable and perceptual qualities that shape spatial experience. By enabling dynamic lighting adaptation, this pipeline equips architects and designers with a visualization tool for analyzing how spaces evolve under changing conditions.
References
[1] J. Gao, C. Gu, Y. Lin, H. Zhu, X. Cao, L. Zhang, and Y. Yao, “Relightable 3D Gaussian: Real-time Point Cloud Relighting with BRDF Decomposition and Ray Tracing,” arXiv preprint arXiv:2311.16043, 2023.
.png)
Figure 2: From 2D to 4D reconstruction
Pipeline
This research proposes the development of a more flexible tool for 3D reconstruction of architectural spaces. This tool will leverage Gaussian Splat techniques to create models that can adapt in real-time to various lighting conditions using just a minimal set of images captured under a single lighting condition. The aim is to enhance the realism of 4D architectural models, which incorporate the dimension of time alongside spatial dimensions, allowing for dynamic interactions with environmental changes.
The technical part of this research is based upon the model in this paper: Relightable 3D Gaussian: Real-time Point Cloud Relighting with BRDF Decomposition and Ray Tracing[1]. This work innovates on traditional 3D Gaussian techniques by incorporating additional properties such as normal vectors, BRDF parameters, and incident lighting from various directions. These enhancements facilitate the accurate recovery of geometry and materials through inverse rendering techniques, address complex occlusions with point-based ray tracing, and enable dynamic relighting of the scene.
Figure 3: Pipeline
Implementation
These relighting experiments under various environmental conditions illustrated the model's robust capability to adapt to dynamic lighting changes.
Figure 5: Relighting results
.png)
An user interface that allows real-time visualization and interaction with the 3D Gaussian Splat model is further developed. This interface enables users to manipulate the model—dragging, scaling, moving, rotating—and observe how it responds to different lighting conditions, ultimately achieving a dynamic 4D model.
Figure 6: Interactive GUI
User Study
This part is a work in progress.
The technical workflow comprises:
Data Collection: Acquiring multi-view images from an architectural site.
Data Processing: Converting images into RGB, depth maps, normal maps, and extracting camera parameters.
Model Training: Training a Gaussian Splat representation from captured data.
Relighting: Applying ray tracing techniques to simulate lighting variations.
Visualization – Implementing a real-time rendering GUI.
Using a dataset of 300 images of a barn, the reconstruction process was completed in 20 minutes on an RTX 4090 GPU. The resulting Gaussian splat representation encoded key surface attributes such as color, opacity, depth, and normal vectors, enabling precise material rendering. Once reconstructed, the model undergoes relighting via environment maps.
Figure 4: 3DGS reconstruction result
