Google taught A.I. to make pro-level photos out of Street View panoramas — check out the results
- To do their experiment, Google researchers worked with around 40,000 Google Street View panoramas, most of which were captured in nature areas like the Grand Canyon and Yellowstone National Park.
- Google had real professional photographers evaluate the images the system created, and more than 40 percent of evaluated photos were rated at a semi-pro level.
Researchers at Alphabet subsidiary Google have taught software to create great-looking photos out of panoramas that you can look at in Google Street View.
An academic paper published on Wednesday documents the Creatism system, which is just the latest fascinating thing Alphabet is doing with artificial intelligence -- alongside Facebook, Google and other tech companies. Other novel Alphabet efforts include the DeepDream image enhancement software and the NSynth synthesizer.
While Google has not put the Creatism technology into production, in time it could be used to improve what's available publicly. After all, today Google Street View panoramas are available in Google Maps, an app that has more than 1 billion monthly active users.
Besides cropping a section of a panorama, the system also adjusts the saturation, high dynamic range (HDR) and light in order to get something that looks nice -- just as a human photographer might do after taking an original picture. "Experiments show that a portion of our robot's creation can be confused with professional work," Googlers Hui Fang and Meng Zhang wrote in the paper.
See what you think -- here are some of the creations, above the panoramas they're derived from:
More Creatism images and their original source images can be found here.