Sex-based approach to estimate human body fat percentage from 2D camera images with deep learning and machine learning

Obesity is one of the most concerning nutritional issues since it is a significant risk factor for chronic diseases, including cardiovascular disease and diabetes. Many dietary disorders require an anthropometry assessment and body fat percentage (BFP) information. Dual-energy X-ray absorptiometry (DXA) is the most precise and automated method for determining BFP; nevertheless, it is costly and difficult to locate in clinics. This paper proposes the utilization of digital image processing and machine learning techniques to estimate BFP, considering four 2D camera images and additional factors such as age, weight, height, and sex. Our proposal specifically adopts a sex-specific approach. Our experiments included pre-processing steps and several regressors. Moreover, we built a dataset composed of 912 samples, including male and female individuals. The sex-based approach to estimating the BFP achieved satisfactory results for both males and females. Thus, it can assist monitor patients as a mobile application, especially in areas where experts and technology, such as equipment, are scarce.