Counting calories is a notoriously inexact science. Even many nutritional labels get their numbers wrong. But as with many things in life, Google is here to help. The tech giant is working on new technology that would be able to count calories in photos of food. It’s like Google Translate for how gross of a person you are.
Google researcher Kevin P. Murphy revealed the project during the Rework Deep Learning Summit in Boston. Called Im2Calories, the program uses deep learning algorithms, a type of artificial intelligence, to analyze photos. Supposedly, resolution is no obstacle: It can even work for Instagram pics.
According to Popular Science, Im2Calories managed to count two eggs, two pancakes and three strips of bacon in a photo, then compared the size of the food items to the size of the plate and assessed any condiments in order to spit out a final calorie estimate.
Like many of these new AI projects, the algorithm isn’t perfect: It will continue to build over time as it collects more data. “We semi-automate,” Murphy said, noting that users could correct certain info if need be. “If it only works 30 percent of the time, it's enough that people will start using it, we'll collect data, and it'll get better over time.” The other potential hiccup comes from actual calorie counting. The technology can only tell what foods are included in the photo. It doesn’t have some sort of magic calorie sensor, so it has to rely on publicly available numbers.
Still, it’s intriguing technology that could have broader implications for the food world and beyond. Murphy sees Im2Calories as much more than a gimmicky new way to play with food porn pictures. “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.”