Lose weight until next summer: a neural network was taught to evaluate the calorie content of dishes from a photo

Startup SnapCalorie launched app of the same name that uses AI to calculate calories food from a photo from a smartphone. Unlike analogues, it also knows how to determine the portion size and quality of the dish. The service is currently only available in the US.

There is a lot of skepticism in the healthcare industry about such tools. And there are good reasons for that. One of research 2020, in which experts compared the most popular calorie counters, showed that the most accurate - Calorie Mama - was correct only about 63% of the time.

SnapCalorie developers claim that their neural network is much more accurate. This is due to a "unique training data set" of 5,000 ingredients, including sauces and oils.

He acknowledges that the algorithm may be biased towards American food, since the team collected most of the original data in the United States. But he was also taught on the dishes of other countries.

In addition, SnapCalorie has a built-in ChatGPT-based chatbot that provides food recommendations based on the user's goals and past preferences. In addition, the nutrition log from the app can be synced with other health tracking platforms such as Apple Health.

SnapCalorie was founded by former Google employee Wade Norris and aerospace systems engineer Scott Baron. The startup is engaged in developments in the field of healthcare.