NutriSight
Overview
- Nutritional information is crucial in assessing the quality of food products, but often, this data is presented to consumers in the form of nutritional tables, which can be challenging to comprehend.
- This project aims to address this issue by developing a tool that can automatically extract nutritional information from a photograph of food product’s packaging. Leveraging deep learning techniques, the solution will encompass a machine learning model and an associated API, capable of operating in a highly multilingual context, enabling the extraction of nutritional data from photos across a wide spectrum of countries and languages.
What it is
NutriSight projects aims at developing a tool that can automatically extract nutrition information from a photograph of food product’s packaging using deep learning techniques.
Project information
This project is funded by the European Union’s Horizon Europe research and innovation action program through DRG4Food open call.
Status: in progress
Start date: May 2024
End date: April 2025
Project lead: Raphaël B
Project notes: https://docs.google.com/document/d/1kBcjxScTOim4wU1VRVW7WOP0Jo1plLQU3xV5-j55MyM/edit?usp=sharing
News
Key dates and numbers
Why it matters
- Adding nutritional information to a product is time-consuming and error-prone for Open Food Facts contributors. A machine learning model would save contributor time by automating the extraction process and would detect possible errors in nutrition values.
- Besides, 73 000 products have a nutrition photo selected but no nutrition information available: this model would allow to extract this data automatically.
What we’ve done recently
- available in the Open Food Facts mobile app
- available as a Web Component
- available in the Dart package for Flutter app
- available on the Open Food Facts website
What we are currently working on
- refinements in the mobile app and website
What’s coming up next
- improving speed of model
- improving to more nutrients and more languages