Product design and development firm Cambridge Consultants developed a deep learning-based system that turns human sketches into paintings that resemble Van Gogh, Cézanne and Picasso.
“What we’ve built would have been unthinkable to the original deep learning pioneers,” said Monty Barlow, director machine learning at Cambridge Consultants in reference to their interactive system that call Vincent. “By successfully combining different machine learning approaches, such as adversarial training, perceptual loss, and end-to-end training of stacked networks, we’ve created something hugely interactive, taking the germ of a sketched idea and allowing the history of human art to run with it.”
Using an NVIDIA DGX-1, they trained their Generative Adversarial Networks (GANs) on 8,000 paintings from the Renaissance period to the current day to study the works and build an understanding of where contrast, color and texture. Once trained on nearly 200 million parameters, Vincent is able to understand the important edges in paintings and uses this understanding to produce a complete picture.
“We’re exploring completely uncharted territory – much of what makes Vincent tick was not known to the machine learning community just a year ago” said Barlow. “We’re excited to be at the leading edge of an emerging, transformative industry and to be making the leap from the art of the possible to delivering practical machine learning solutions for our clients.”