The “FoodRec” project was born from the need to address the growing epidemic of obesity and its chronic complications, proposing an objective, fast, and automated approach to recognize both quantity and type of food using the smartphone camera. This work is part of a research line that integrates artificial intelligence, mobile health, and automated monitoring of eating behaviors, aiming to support users in the conscious management of their daily diet.
Motivations & Objectives
- Go beyond the unreliability of questionnaire or self-report-based methods using AI systems for direct food recognition from images.
- Rely on accessible technology (smartphones) manageable by anyone to support adoption and solution scalability.
- Continuously monitor both caloric intake and energy expenditure, integrating data from wearable or mobile sensors.
Methods and Technologies
- Integration with a mobile app for semi-automatic food diary collection and visualization, enabling personalized caloric and dietary feedback (ActiveMiles app).
Results, Applications & Impact
- The FoodRec architecture has outperformed baseline models in the recognition of food in real images collected by app users during personalized clinical pathways (e.g., smoking cessation programs).
- The use of a user-bias branch in the CNN enables learning and monitoring of individual eating habits, tailoring suggestions and feedback in a personalized way.
- The user-friendly interface and app portability facilitate monitoring and timely intervention on unhealthy habits, representing a tool for primary prevention.
Related Scientific Articles
- ActiveMiles: an intelligent food intake and energy expenditure mobile application
Presentation and technical explanation
AI- and computer-vision-based app that recognizes food, monitors activities, and estimates calorie consumption in real time. - Real-Time Food Intake Classification and Energy Expenditure Estimation on a Mobile Device
(Body Sensor Network Conference, 2015)
Proposes a wearable system for real-time automatic food intake classification and energy expenditure estimation.
Code Repository
- FoodRec App, dataset and models: code available upon request for cooperation projects and scientific testing. For specific requests, contact the authors or Daniele Ravì on GitHub.
Team & Authors
- Daniele Ravì (concept, AI algorithm development, HealthLab/Imperial-UCL project supervision)
Interdisciplinary project between UK and Italy universities – Tested on real trials and clinical support.