Body Weight Estimation using 2D Body Image

Abstract

Two dimensional images of a person implicitly contain several useful biometric information such as gender, iris colour, weight, etc. Among them, body weight is a useful metric for a number of usecases such as forensics, fitness and health analysis, airport dynamic luggage allowance, etc. Most current solutions for body weight estimation from images make use of additional apparatus like depth sensors and thermal cameras along with predefined features such as gender and height which generally make them more computationally intensive. Motivated by the need to provide a time and cost efficient solution, a novel computer-vision based method for body weight estimation using only 2D images of people is proposed. Considering the anthropometric features from the two most common types of images, facial and full body, facial landmark measurements and body joint measurements are used in deep learning and XG boost regression models to estimate the person’s body weight. The results obtained, though comparable to previous approaches, perform much faster due to the reduced complexities of the proposed models, with facial models performing better than full body models.