Exploring AI-Based System for African Food Weight-Loss Recommendations

Abstract

Listen

Translate article

This research leverages artificial intelligence to design an African food recommendation system for weight loss. The rationale for designing this system was based on our recently published study on the design of socio-cultural food recognition systems for Africans. Based on our previous study, results revealed that users considered the socio-cultural food recognition system to provide nutritional value and would require a robust system with more African foods. Hence, to tailor our findings to effective dietary planning where obesity could be a concern, we propose the current system given the health implications of additional foods for specific users (that is, overweight users). Our current study is in three phases. The first phase will focus on validating some African foods with dieticians to determine their appropriateness for weight loss and better alternatives based on calories and other important metrics. Additionally, we will invite dieticians and some overweight users to evaluate some low-fidelity (Lo-fi) prototypes for the design requirement elicitation of the final prototype. The second phase will involve the development of our AI models (computer vision and large language models) and their evaluation. Furthermore, we will leverage the design requirements gathered from the lo-fi prototype study together with the AI models to develop a high-fidelity (Hi-fi) AI system that will run on mobile devices (the final prototype). Consequently, a post-study evaluation will be conducted with dieticians and overweight users to obtain subjective feedback. Hence, findings from this study will provide design recommendations for integrating African foods into existing and related large-scale AI-based systems in the future.