EAGER: Exploring the Relation between BMI and Visual Appearance of Face and Body

身体质量指数（BMI）是衡量个体体重与身高之比的指标，是将人体分为四类的重要参数，即，体重不足、正常、超重和肥胖。高BMI值与2型糖尿病、高血压、心血管疾病和某些癌症等疾病的风险较高相关。本项目探讨BMI与人脸和身体视觉外观之间的内在和基本关系。肥胖症流行的一个主要原因是许多人不知道他们的BMI以及与高BMI值相关的各种疾病的高风险。该项目的研究导致了一个智能和计算系统的开发，该系统可以被任何人在任何地方和任何时间使用。开发的系统使人们能够了解他们的BMI，并了解各种疾病的强烈相关高风险，以对抗超重和肥胖。本研究探讨并确定BMI与面部和身体的视觉外观之间的关系，为未来开发一种低成本、便携、可靠、方便的BMI评估系统提供知识。研究的关键问题是从人脸和人体图像中提取什么样的特征或模式来表征与BMI测量相关的视觉外观。在人脸图像中，重点研究了二维人脸特征的表示及其鲁棒性，以建立人脸到BMI的映射关系。在身体图像中，重点是3D体型分析，以连接到BMI测量。该研究为开发一个可视化分析系统提供了理论基础，该系统可以随时随地提供方便的BMI和相关健康状况的估计。

Body mass index (BMI) is a measure of the ratio between an individual's weight and height, which is an important parameter to characterize human bodies into four categories, i.e., underweight, normal, overweight, and obese. A high BMI value is associated with a higher risk for conditions such as type 2 diabetes, high blood pressure, cardiovascular disease, and certain cancers. This project explores the inherent and fundamental relation between BMI and the visual appearance of human face and body. A major reason of the prevalence of obesity is that many people are not aware of their BMI and the higher risks of various diseases associated with high BMI values. The study of this project leads to develop-ment of an intelligent and computational system that can be used by everybody at anywhere and anytime. The developed system enables people to aware their BMI and understand the strongly correlated high risks of various diseases to combat overweight and obesity. The developed technology can also improve personal health care and quality of life, and public health surveillance.This research explores and determines the relation between BMI and the visual appearance of face and body, in order to obtain knowledge for future developing of a low-cost, portable, reliable, and convenient BMI assessment system. The key research question is what kinds of features or patterns can be extracted from human face and body images to characterize the visual appearance related to BMI measure. In facial images, the study focuses on 2D facial feature representation and its robustness in order to build the mapping relation from face to BMI. In body images, the focuses are on 3D body shape analysis to connect to BMI measure. This research provides a theoretic foundation for developing a visual analysis system that can be deployed to provide convenient estimate of the BMI and related health conditions anywhere and anytime.