AI-Powered PicsArt Magic Effects Coming to Smartphone Near You | NVIDIA Blog
October 5, 2016
blogs.nvidia.com · by
House or horse, bird or barn? Deep learning and GPU computing have quickly advanced the abilities of image recognition technology to superhuman levels.
Now, PicsArt, maker of the social photo editor by the same name, is applying this breakthrough in artificial intelligence to the creation of images.
Hitting the market today, “Magic Effects” is a new feature in the latest version of the PicsArt app, which has been downloaded more than 300 million times and boasts 80 million active monthly users. Magic Effects uses GPU-powered AI to analyze the quality and context of photos, and enables users to transform their pics in seconds with an array of filtering effects that are customized based on the AI analysis.
If, for example, a user applies the “Neo Pop” effect to a photo, the result won’t be standardized. Instead, it will be customized based on the qualities of the photo in question. Users can further customize the filters using a variety of touch-interface tools.
Here’s an example of a Magic Effects filter in action, turning a photo into a colorful painting.
Making Everyone an Artist
The idea is make art and photography more accessible to a larger number of people, says PicsArt co-founder Artavazd Mehrabyan.
“With AI, we can build much more sophisticated tools that understand the visual context of a photo and apply learned techniques to make the image more of what the photographer wants,” said Mehrabyan. “Artificial intelligence is opening a whole universe of new creative techniques in photography.”
The updated PicsArt app also will enable users to search for similar images, search by image type and generally allow amateur photographers to do more with their photos than ever before.
GPUs Meeting Demands of AI
PicsArt used a combination of GeForce GTX GPUs and Amazon Web Services GPU instances to train its AI and similar-image search algorithms, as well as for scaling and production. Mehrabyan said the company could not have used CPUs to achieve the kind of speed to market and photo effects capabilities that GPUs enabled.
“Building the AI models behind our new Magic Effects requires long, multiple day training sessions for each filter,” he said. “This is a process we have to repeat over and over again until we get optimal results.”
The capabilities PicsArt is introducing are only the beginning. GPU-powered AI will continue to expand the possibilities in photography, enabling anyone with a smartphone to turn their photos into works of art, says Mehrabyan. (And share it via social media, which PicsArt makes easy.)
“We see much more automation on the photo tools side, and that’s going to make it possible for the mass consumer to have easy access to the most advanced creativity tools,” said Mehrabyan. “AI will not replace creativity, but will enable people to explore and express their creativity.”