In work exhibited at the NeurIPS 2017 Machine Learning for Creativity and Design, we developed a process for relating depthmaps extracted from Google Street View panoramas with the corresponding photographic information. A number of separate models were produced using limited geographic areas of selected cites. With these depthmap-to-panoramic cityscape models trained, we are able to generate new images from unrelated depthmaps which resembled photographic images of the selected cities. This is demonstrated using a javascript app (not currently online) and documented in still images and a series of videos.

In work exhibited at the NeurIPS 2018 Machine Learning for Creativity and Design, three-dimensional architectural massings for single-family homes are generated by a generative adversarial network. This GAN is trained on a small dataset of three-dimensional models of homes falling into seventeen architectural styles, and that are represented as multi-view heightfield images. A process is developed for converting from 3d CAD model to 2d tiled heightfield image, and from the 2d heightfield images generated by GAN back to three-dimensions in voxel format.