Siri Creator Builds Platform That Guesses Calorie Counts From Photos
The developers behind the iPhone feature are developing an image recognition technology that gives users an approximate calorie count of their food.
- 4 april 2014
SRI InternationaI, the developers of the technology behind Siri, are developing a new method that uses image recognition to give an approximate calorie count of meals by analyzing photos of the food.
The idea behind the project, internally called Ceres by SRI, is similar to what the app Meal Snap does, which is tell users what was in their meal through snapshots. While Meal Snap uses a camera and Amazon’s Mechanical Turk service, SRI’s project will let computers estimate the calorie count. To do that, SRI had to solve two problems – identifying the food and estimating the volume of the food.
According to SRI Ventures executive director Dror Oren, it’s difficult to get an accurate count of calories in food because of hidden fats such as oils and butter, but he explained to Gigaom that providing a close range of the calorie count is already useful for many people.
The main technology used is image recognition, but the project also plans to use other information as context clues to get a more accurate calorie count. Information such as where the user is having the meal or information taken from other food apps can be used so Ceres can ask the right questions and arrive at an accurate count. The image processing and recognition will all be done in the cloud.
SRI is still looking for more partners for the project and expects the methodology to be released in the mainstream in about a year.