Special Seminar in Computer Graphics

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

Filip Jurčák
Material Picker: Material recognition in images using machine learning

One of the important steps in modeling realistic 3D scenes is setting material appearance of the various scene objects. The goal of our ongoing project is to simplify this often tedious task by providing the 3D artist with an intelligent material picker tool. The tool should allow user to ‘pick’ a material from any given input image by simply pointing to an object. A deep neural network will be trained to achieve this nontrivial goal. An extensive set of training data will be provided, where the complex correspondence between the image pixels and the underlying object material will be available. The network should be then able to recover this pixel-material correspondence from new, previously unseen images.