Facial Image-Based BMI Estimation Using ResNet50 for Obesity Screening

Public health challenges associated with obesity include diabetes and cardiovascular disease, which require effective screening methods for early intervention. Body Mass Index(BMI) is a soft biometric measure to evaluate obesityrelated risk factors. We can determine a person’s adiposity by measuring BMI from facial images. This study uses deep learning techniques to determine the feasibility of estimating BMI from facial image analysis. We developeda convolutional neural network based on ResNet50 architecture to detect facial features. The neural network uses transfer learning and weights that have been pre-trained. BMI values are calculated by analyzing anthropometric parameters, such as cheekbone width, jaw width, and facial height ratios. Based on these findings, facial image analysis may be a practical alternative metric in health monitoring for detecting obesity-related risks early and easing prompt interventions to improve health outcomes.