Estimation of Body Mass Index from Facial Image Using Modified ResNet and Ridge Regression

Obesity stands out as a critical global public health concern, prompting efforts to assess and monitor individuals’ health through the use of Body Mass Index (BMI). A commonly used statistic, BMI is calculated by dividing a person's weight by the square of their height. Despite its significance, determining BMI from facial images poses a considerable uncertainty in the realm of artificial-intelligence technology, leading investigators to explore different approaches. This study researches a brand-new technique by utilizing facial photos to estimate body mass index (BMI). Based on our research we conclude that ResNet-50 model gives the best results compared to all the deep learning models. So, our research mainly focuses on customizing the ResNet-50 model while trying with different parameters and finding the best result. The method predicts an individual's BMI by first extracting data from facial photos using customized ResNet-50 model and feeding it to a Ridge Regression model. For this objective, a dataset of 65,000 facial photographs was gathered, comprising both male and female subjects with varying BMI ranges.