Automated food intake calculation

The main reason that motivates hospital managers to install this type of device is to help tackle undernutrition among the elderly.

In addition, other motivations lead companies to equip themselves with our system:

Monitor each patient's food intake in terms of overall quantity or nutrients/energy per meal/day/week/stay
Be alerted when a patient's consumption evolves negatively, or if its variation needs to be observed.
Take charge of the diagnosis of patients at risk, identify and maintain control of the diagnosis and actions.
Diagnose the situation of the service with regard to food waste.


A public/private partnership initiated by the CHU of Dijon, which brings the profession, piloted and implemented by Atol CD with the expertise and innovation of Yumain in AI.

• The Geriatric Research Unit of the Elderly Pole of the University Hospital Center (CHU) Dijon Bourgogne is a structure dedicated to the knowledge of the nutrition of the elderly.

• Atol Conseils et Développements is specialized in the development of custom web and mobile business applications and is a recognized open source integrator.

• Yumain is a specialist in intelligent recognition solutions and advanced industrial cameras integrating embedded artificial intelligence (Edge-computing).

Atol (5)
Atol (4)


1. Picture taking

A mobile device attached to the meal distribution cart recognizes the presence of the tray and automatically takes a picture before and a picture after the patient has eaten his meal.

2. At the end of the tour, the photos are sent to the AI

AI accurately identifies each ingredient on the BEFORE tray
AI accurately identifies each ingredient on the AFTER tray
The AI compares volumes of food Before/After = automated calculation of the ingesta

3. Data transmission

Via a trigger on the tray, the data from the information system (restaurant, diet, DPI) are collected and aggregated with the meal consumption values calculated by the AI.

4. Monitoring of ingesta

The follow-up of the ingesta and nutritional intake over time is provided for each patient. A nutritional alert engine can be configured to warn dieticians of a potential need for intervention.


The device for detection and identification of ingredients and their volumes on a tray has an accuracy of over 90%. This device is based on the combination of our robust algorithms and filters applied by image processing. The neural network learns every day from the new images it receives.


The artificial intelligence analyzes a couple of before-and-after trays in less than 2 seconds.


Visit the dedicated website Foodintech (in French)

Ask for a free demo !

We will contact you immediately.