LightMirror: AI-driven renovation planning
LightMirror is one of seven winners of the second ELSA Industry Call and has received 60,000 euros in funding. For the duration of the Industry Call, LightMirror has worked together with top AI experts within our Network of Excellence. LightMirror is dedicated to pioneering new ways to harness sunlight and accelerate building renovation for greater energy efficiency and comfort.
Result Video
In this video, the team presents the joint project with ELSA:
Short Summary
LightMirror’s short summary of the project:
“During our collaboration with ELSA (January-June 2025), LightMirror developed and validated a high-performance metamodel for predicting building heating needs using artificial intelligence. Our team successfully created a comprehensive synthetic database of 270,000 building instances characterized by 44 thermophysical descriptors and trained Multi-Layer Perceptron neural networks achieving remarkable accuracy with NRMSE of 0.038 and R² of 0.9986. The solution delivers inference times 5,000 times faster than traditional Building Energy Simulation (BES) software. We successfully integrated Blender modeling software with our Python-based AI engine, enabling automated processing of building models for city-scale applications. The project resulted in a scientific publication titled “High Performance Metamodel for Cuboid Buildings Heating Needs Prediction” and participation in major industry events including Vivatech, ELSA General Assembly, Technorama 2025, and GovTech4All bootcamp.