Measurement of Plate Leftovers. A photo analysis using AI.
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: artificial intelligence, food waste, measurement
Monday 6 October 2 p.m. - 3:40 p.m. (Europe/Amsterdam)
Abstract
Assessment of the amount of household food waste is challenging, and prior research has mainly used surveys, diaries, or bin composition analyses to measure this amount of waste. Yet, all these methods have sizeable barriers and issues related to representativeness, completeness, validity and/or measurement error. The coding of photographs of food waste has been proposed as a promising alternative, but manual coding of these photos is cumbersome both in terms of researcher time and effort. The current study shows how recent artificial intelligence (AI) advances can be used to approximate the leftovers on dinner plates. Based on 468 photos of dinner plates with leftovers, we show how GPT4vision can be used to obtain a reliable estimate of the type and quantity of the visible remnants on plates. We examine different user and system prompt messages, to provide insights on the best approach. Use of this approach would severely decrease the effort associated with photo coding, making photograph-based measurement of food waste a readily accessible method for food waste scholars.