Recompose: A Real-Time Vision and Feedback System to Reduce Food Waste

Umang Sharma

Abstract: Food waste is a major sustainability challenge, impacting the environment, economy, and broader society. Schools and cafeterias are often hotspots for wasted food, making them ideal places to test solutions. This study explores how real-time feedback, delivered through technology, can reduce food waste while encouraging sustainable behavior. We developed Recompose, a "cognitive composting" system that combines computer vision with insights from behavioral science. Using an Intel RealSense depth camera, Meta?s Llama Vision AI, and a Raspberry Pi, the system identifies, measures, and classifies discarded food while estimating its environmental impact, including greenhouse gas emissions, water use, and energy consumption. Personalized feedback is provided immediately at the point of disposal to guide behavior. In a five-day high school trial, daily food waste dropped by 36.9%, and results suggested early signs of habit formation and sustained awareness. These findings indicate that affordable, scalable technology paired with thoughtful behavioral design can not only reduce food waste but also foster lasting sustainable practices, offering a practical model for schools and other institutional settings.

Keywords: food waste, computer vision, behavioral intervention, real-time feedback, sustainability

How to Cite?: Umang Sharma, "Recompose: A Real-Time Vision and Feedback System to Reduce Food Waste", Volume 14 Issue 10, October 2025, International Journal of Science and Research (IJSR), Pages: 210-213, https://www.ijsr.net/getabstract.php?paperid=SR251001091332, DOI: https://dx.doi.org/10.21275/SR251001091332