Applying deep learning image recognition technology to promote environmentally sustainable behavior

Abstract

Since the dawn of the 21st century, the carbon footprint has become an important indicator for measuring environmental impact. While it's been recognized that changing one's dietary habits can help to reduce food-related carbon emissions, many consumers do not yet understand much about carbon emissions generated by food production and dissemination and do not actively search for related information. Thus, developing a way for consumers to identify a food's carbon footprint easily and quickly and perhaps to then adjust their eating habits is one way to help achieve environmentally sustainable eating behavior. This study establishes a carbon footprint tracking system using convolutional neural network image recognition technology to explore and improve the advantages of environmentally sustainable eating behavior. Results show the accuracy of the proposed model is 94.79%, indicating it can effectively recognize food types. During a two-week experiment testing the tracking system, the measured total carbon footprint for the study participants was reduced by 22.25%. Additionally, the number of participants whose daily carbon footprint was less than the targeted goal of 3 kg increased by 81.82%, indicating that the system can effectively help people reduce their food carbon footprints while providing them with calorie and nutrient information required to help achieve a balance between health and environment.