nutrItioN healTh foodsErvice iNdustry dieT ImprOvemeNt

Current approaches to food waste and dietary intake assessment in collective catering mainly rely on manual methods, such as direct observation, food diaries, and weighing procedures, which are time-consuming, costly, and prone to bias. INTENTION advances the state of the art by introducing an automated, AI-based approach using computer vision models (Mask R-CNN with weight regression) applied to food images. The system reduces manual data collection, improves objectivity, and enables scalable and continuous monitoring through integration with a WebApp and a structured data infrastructure designed according to privacy-by-design principles.

The project was structured into five Work Packages (WP1–WP5), all completed successfully.

• WP1 focused on the systematic review, development of the image acquisition system (BolC), the Mang-IA WebApp, and training and validation of computer vision models.
• WP2 addressed the definition of real-world validation protocols and recruitment of participating canteens and facilities.
• WP3 involved validation in real settings (hospital and workplace canteens), data collection and performance assessment.
• WP4 covered communication and dissemination activities, including national events, webinars, scientific posters, and online materials.
• WP5 ensured technical, administrative and financial coordination and reporting throughout the project.

The project successfully developed and validated an integrated AI-based system for automated monitoring of food consumption and waste in collective catering settings. The solution has been demonstrated in relevant operational environments, reaching an estimated TRL 5–6. The system enables scalable and data-driven assessment of food intake and leftovers, supporting menu optimization, portion sizing, and sustainability strategies. Results confirm the feasibility and applicability of the approach and provide a strong basis for further upscaling and wider deployment.