DaVinciFace laverages the Deep Learning technology by using a convolutional neural network, trained on the DaVinci’s most famous portraits such as “La Gioconda” or “La Belle Ferronière”, with more than 500 million parameters.

More in detail, the convolutional architecture of the generative neural network (GAN) is structured on 56 layers with about 10,000 (compound) parameters each. The source image (photo) is processed by extracting its main features, compressing them in a latent space, and then regenerating the final image by using the features extracted from the paintings of Leonardo DaVinci.

Given the high computational capacity required to complete the processing in a reasonable time (now about 100 seconds), the application relies on the graphics hardware currently available on the state-of-the-art graphic hardware.