X

Google's 'Im2Calories' AI To Count Food Photo Calories

Counting Calories can be a useful tool for those who are trying to watch their weight, stick to a strict eating regimen as part of a workout routine, or for anyone who is just generally curious what their calorie intake might be. For those who are consistently taking pictures of their food and sharing it via Instagram and other social media sites, (guilty) having your calories counted for you by an intelligent computer program or algorithm could be an infinitely useful tool if they also fall into any group mentioned above. Google is reportedly working on an AI (artificial intelligence) project which is aiming to do just that, count the calories for each piece of food that people take photos of.

The project is called Im2Calories, and if Google is able to perfect this project it will be able to identify any food items that people capture in photos and then associate a calorie amount to each item, using visual analysis and pattern recognition. At current Google’s progress on this new AI is still going through extended research, but like many other learning applications or services which people may use on a daily basis, (take third-party keyboards on an Android device for example which learn your vocabulary over time) Google states the Im2Calories tech will work in such a way that it learns the ability to read the correct caloric intake of individual food items in pictures over time, with the more pictures are taken of food and the more they’re shared, by grabbing caloric data already found on the web and piecing it together like a puzzle with the images of food.

It sounds like a lofty goal but if there’s any company capable of developing such a technology, Google would be it. According to Kevin P. Murphy who works with Google’s research lab, Im2Calories uses “sophisticated deep learning algorithms” to calculate the correct data for calories and apply it to photos. Past being able to give a calorie count to each and every one of your instagram food porn shots though, Murphy states that hopefully the technology could also be used and applied to other situations like analyzing traffic patterns to help people find parking spots. When and if the AI reaches this point is still an unknown but the research and the technology behind are interesting nonetheless.