Google is developing an application associated with artificial intelligence to identify the elements that constitute a dish, simply from a photograph. The idea would thus be to indicate to the user how many calories the dish he has photographed contains, with the objective of helping him to eat better.
Internet users have been increasing the number of photos of food for a few years, and social networks have been dealing with the "foodporn" phenomenon. Nothing too bad in itself, but it gave Google ideas.
The Im2Calories project could thus precisely identify the food captured in a photo and give indications to the user and in particular the caloric data. The algorithm would be able to identify foods, but also to estimate their weight by taking into account the dimensions of the plate and other elements.
According to one of the researchers in charge of the project, Kevin Murphy, the mobile application would be able to give a result in a few seconds with deviations of around 20% from reality. The development of the application is still in progress and no availability date is official for the moment. Let's just hope that the estimates will be more realistic than some estimates from Microsoft's AI which suggests guessing the age of users in a photo.