Introduction

Currently the main commercial light-planning, design and simulation software in market are the DIALux, AGi32 and Relux products. However, they can simulate the light energy propagation only within full complete closed models and cannot address sparse models or models collected and reconstructed with commodity depth sensors. Moreover, they struggle with increased computational resources and intensive simulation timings. To overcome these shortcomings, the inventors propose the use of data driven solution for faster and reliable light-planning, design and simulations without compromising the photometric accuracy.

Technical features

DeepLux is a data driven pipeline for predicting illuminance maps, i.e., light estimation, on the 3D space. Unlike existing approaches, the presented method is the first learning-based approach that is able to predict accurate illuminance maps in real time. Such maps are fundamental for applications such as environmental light modelling and design, as well as for smart lighting systems that can actively change light intensities given the presence/absence and activities of humans in indoor environments. DeepLux can lead to a new generation of lighting management systems targeting for high quality light, power efficiency, safety and well-being.

Possible Applications

• Light management systems;
• Energy saving, safety and well-being;

Advantages

• Learning-based data driven model;
• Increased prediction accuracy;
• Real time light estimation and simulation.