AI to the Rescue: A Novel Approach to Measure Plate Leftovers
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Assessment of household food waste has mainly relied on surveys, diaries, or waste bin composition analyses, each challenged by representativeness, completeness, validity, and/or measurement error. Photographing food waste is a promising alternative but manual coding requires time and effort. We demonstrate how recent advances in artificial intelligence (AI) can be used to analyze photographed leftovers. Based on 816 photos of leftover dinner plates, we use GPT4o to obtain a reliable estimate of the type and quantity of visible remnants on plates. Based on our prompt development, we provide insights on how to use AI. We validate our findings, and assess the impact of measurement error by respondents, with 93 pictures collected by experts. The use of this novel approach severely reduces the effort associated with photo coding, while maintaining high precision and allowing for easy up-scaling.
