Original title: PatchBMI-Net: Lightweight Facial Patch-based Ensemble for BMI Prediction
Authors: Parshuram N. Aarotale, Twyla Hill, Ajita Rattani
In an article focusing on the alarming rate of obesity, researchers have developed a new method to predict body mass index (BMI) using facial images. Previous studies have used complex convolutional neural network models for this prediction, but their computational requirements make them unsuitable for resource-constrained mobile devices like smartphones. The researchers propose a lightweight facial patch-based ensemble called PatchBMI-Net specifically designed for BMI prediction and weight monitoring using smartphones. Through extensive experiments on BMI-annotated facial image datasets, they found that PatchBMI-Net achieved similar accuracy to heavyweight models like ResNet-50 and Xception, but with a significantly smaller model size and faster inference time when deployed on an Apple-14 smartphone. This innovative approach offers an efficient and low-latency solution for on-device weight monitoring using smartphone applications, which could greatly benefit the 93.3 million adults affected by obesity in the United States.
Original article: https://arxiv.org/abs/2311.18102