Neural Acceleration of Scattering‐Aware Color 3D Printing

With the wider availability of full-color 3D printers, color-accurate 3D-print preparation has received increased attention. A key challenge lies in inherent translucency commonly used print materials that blurs out details color texture. Previous work tries to compensate for these scattering effects through strategic assignment colored primary printer voxels. To date, highest-quality approach uses iterative optimization relies on computationally expensive Monte Carlo light transport simulation predict surface appearance from subsurface within a given material distribution; optimization, however, takes order days single machine. In our work, we dramatically speed up process by replacing with data-driven approach. Leveraging deep neural network highly heterogeneous medium, method performs around two orders magnitude faster than rendering while yielding results similar quality level. The is based an established atmospheric cloud rendering, adapted domain and extended physically motivated weight sharing scheme substantially reduces size. We analyze its performance end-to-end pipeline compare runtime alternative approaches, demonstrate generalization unseen geometry values. This first time enables full heterogenous frames actual printing time.