Crowd Generation using Morphological Obesity Criteria

Generating virtual humans is a great challenge and require the solution of various problems. One of these is deforming the body to create a crowd population for games or simulations. Many visual and morphological criteria have been used such as height, sex, and age but, body mass criteria which is directly related to obesity (and BMI - Body Mass Index) has been neglected. We propose a method that generates variants of a scanned model with different levels of obesity. Fat tissue spread in various ways over the body and our method comply with this at body segment level as well as over different sides of each segment following the data gathered for children. We also propose a two-fold validation method: quantitative and qualitative. To validate generated models, an analysis of models' geometric volume associated with each level of obesity and a questionnaire was presented to 40 experts on Body Perception. We found that size is manageable to perform and assure. On the other hand, experts opinion although favorable to our approach were inconsistent some times: experts were able to identify the fact that base models were from the lower end of the BMI scale used (probably because their head were not deformed) but they could not spot which model was the original one (perhaps because our approach could have fooled them). The data show that our strategy is be able to generate credible virtual humans on various levels of obesity.