The app MyFoodRepo

MyFoodRepo is an application developed by the Digital Epidemiology Lab at EPFL. This application allows users to easily track their food consumption. The app uses either the barcode or a photograph taked by the user couples with an image recognition algorithm to identify what the consumer is using

  • To associate a barcode to a product, we use the FoodRepo database. It is an open database that contains most of the food products with barcodes sold in Switzerland. For more information, you can visit the website www.foodrepo.org.
  • To determine the content of raw or cooked food, we use machine learning algorithms, a way to process data to discover patterns that can later be used to analyse new data. The data that trains the algorithm in myFoodRepo is photos of food and drink. The image recognition algorithm that is implemented in this application is the current state of the art. By providing images, and by fixing mistakes in the suggestions made by the algorithm, the user contributes directly to its improvement.

The estimated nutritional content, i.e , the energy value and the main nutrients (Protein, Carbohydrates, Dietary fibres and Fat) are calculated with the swiss food composition database.

To accelerate progress in this area, all images are openly available by default (we make this large photographic database available to other researchers). No information about the users is publicly accessible. An option in the application settings enables the user to prevent image sharing.

MyFoodRepo was approved by the EPFL ethical committee in July 2017.

More information