The technique is based on “experimental deep-learning system” which uses machine learning for artistic content creation, according to Google research blog.
Using Google Street View, several landscape panoramas were taken and then, post-processing operations were carried out to create an aesthetically pleasing image.
The research approach relied on a collection of professional quality photos of which the subjective aesthetics were broken down into multiple aspects automatically, each of which is learned individually by the machine.
Google used 40,000 panoramas from areas like the Alps, Banff and Jasper National Parks in Canada, Big Sur in California and Yellowstone National Park to conduct the “Turing-test” like experiment, which was then judged by professional photographers.
The team mixed the machine creations with other photos and about 40 percent of them were rated at “semi-pro” to “pro” levels by the judges.