Google is developing an app to count the calories in your Instagram feed
Google's new AI technology could soon let you know how rich the subjects of your food porn snaps really are
Nutrition isn't typically top of mind for the average Instagram user while taking #foodporn snaps at brunch.
It's not that anyone believes the cheesecake-stuffed french toast in front of them is actually health food, but who has time to think about calories when servers keep blocking the light source and one mimosa came without raspberries in it?
A new technology unveiled by Google last week could soon change this by making the experience of shooting snacks with your smartphone a lot more informative — for better or for worse.
Introducing "Im2Calories": A newly-patented artificial intelligence system concepted by Google to analyze still photos of food and and estimate how many calories are in it.
As Popular Science explains, the prospective tool works by determining the depth of each pixel in an image and recognizing visual patterns linked to existing foods.
"Deep learning algorithms" are then able draw connections between what specific pieces of food look like and the "vast amounts of available caloric data" linked to them.
The amount of food in a picture is determined by its size in relation to things like plates and condiments.
Killer app from <a href="https://twitter.com/googleresearch">@googleresearch</a> : image to calories (sic!) <a href="https://twitter.com/hashtag/DeepLearning?src=hash">#DeepLearning</a> <a href="https://twitter.com/hashtag/reworkDL?src=hash">#reworkDL</a> <a href="http://t.co/FwzJ5d3PK8">pic.twitter.com/FwzJ5d3PK8</a>
—@olexandr
Fortunately for social media food-photographers (or, unfortunately, depending on how you look at it) high-res images aren't required. The technology works just as well for standard-Instagram snaps.
"To me it's obvious that people really want this and this is really useful," said Google research scientist Kevin Murphy while presenting the project at Boston's Rework Deep Learning Summit last week. "Now we can start to potentially join information from multiple people and start to do population level statistics. I have colleagues in epidemiology and public health, and they really want this stuff."
While Murphy did admit that Im2Calories isn't completely accurate, he expects this to change as the AI improves itself through use.
"If it only works 30 percent of the time, it's enough that people will start using it," he said, explaining that end-users can manually correct the AI's errors. "We'll collect data, and it'll get better over time."
How much time is unclear, however, as Google has yet to reveal when an app using this technology will be available to consumers for things like calculating daily food intake or determining serving sizes.
Google spokesman Jason Freidenfelds told CNET Tuesday that the app is still in its research stage and that there are "no actual product plans at this stage."
"We don't have more to say at this point on whether or how this might make its way into some future product(s)," he said, "because we'd really just be speculating."
Doughnut-lovers of Instagram, rejoice.