Face To BMI: Predicting Body Mass Index Using Neural Networks

Abstract

Body Mass Index (BMI), a vital health metric indicating weight relative to height, classifies individuals as underweight, normal weight, overweight, or obese, guiding early interventions for health risks, but traditional methods relying on height and weight measurements are labor-intensive, making automated BMI assessment a scalable and efficient alternative for health analysis and decision-making; leveraging facial features, which reveal correlations between face geometry and BMI, we employed large-scale pretrained models like EfficientNet-B7, Swin-Transfomer, and ResNeSt- 101, trained on diverse datasets such as the Illinois DOC, Height-weight-BMI dataset with Celebrity Faces, to enhance prediction accuracy, scalability, and applications in health monitoring, insurance, and policymaking, thereby enabling informed societal and individual health decisions.