ECE.09R – Wastend

• Daniel Derosier
• Connor Shenberg
• Vince Tian
• Juan Zarco

In response to the pressing issue of food waste in households, our project aimed to develop a privacy focused and user friendly application that leverages localized computer vision and other machine learning models. The application empowers users to effortlessly track the food in their homes while offering recipe recommendations that optimize their food stock utilization. The prevalence of food waste in the United States, with 40 percent of all food being discarded, highlights the urgency of addressing this problem at the household level. The intention of our project was to minimize food waste originating from homes, a significant contributor to landfill space consumption, water and energy wastage, and greenhouse gas emissions. Our application utilizes computer vision to identify ingredients on a receipt, capturing images, and storing data in a database alongside their approximate expiration date and the Food Safety and Inspection Services recommended use-by date. We hope to achieve an ingredient recognition accuracy above 80 percent to minimize unnecessary user intervention. The system minimizes food waste by recommending recipes with a bias for using ingredients that are going to spoil along with user defined parameters such as dietary restrictions and cuisine. In our testing, we were able to achieve a mean square error of .019 on a recipe rating dataset with ratings normalized to values between 0 and 1. With 83 percent of consumers lacking a process to prevent spoilage and 82 percent expressing interest in using software applications for assistance, our project directly addresses a critical gap in current practices that is also desired by the population in question.