LUCE

Applied to industry, spectral Monte Carlo (MC) light transport simulation allows addressing predictive rendering purposes. It is the main challenge to replace physical moke-up with virtual prototypes in aesthetic decision-making. However, virtual prototypes need computational costs related to convergence properties that require long rendering times for noise-free images.

To reduce the expensive computational cost, we want to apply a machine learning algorithm to reduce the convergence time. Hence, we want to include a deep learning algorithm in the rendering engine to reconstruct the noise-free image. That's why we merge the a posteriori and a priori methods to extract more information during all rendering steps.