When following a diet, it is not always easy to assess the calories swallowed. We can have fun weighing each food and referring to a calorie table but it is rather tedious. Researchers fromImperial College of London have developed a artificial intelligence able to count the calories of a meal from a simple photo. There are already applications as Foodvisor who are able to do it, but this AI goes even further, taking into account the leftovers left on the plate. The camera continuously monitors subjects as they eat, so the system determines how much food is actually swallowed rather than based on the size of the dish being served.
AI is thus proving to be more reliable than humans in estimating the number of calories in a meal, with an error rate of 37.6% compared to 48.8%, according to the team’s results published on the server arXiv. This AI could not only facilitate diets, but also to improve the quality of research in nutrition: indeed, most studies are carried out on the self-evaluation of the food ingested by the participants, which often turns out to be unreliable and leads to wrong conclusions.