Скачать или смотреть 3DBodyTech2025 - Creation of Avatars in Sportive Poses through a Digital Human Model for Skin Shape

Creation of Avatars in Sportive Poses through a Digital Human Model for Skin Shape Prediction - 25.53

Extract of Presentation of Alejandra POLANCO, Decathlon Sportslab at 3DBODY.TECH 2025
Extended abstract at https://proc.3dbody.tech/abstracts/20...
Full video available in the proceedings

Abstract:
Accurately predicting how skin shape deforms during sportive movements is essential for developing better-fitting ergonomic products, from high-performance apparel to athletic equipment. A digital human model (DHM) with an anatomically parametrized skeleton was developed to combine the potential of skin shape reconstruction and movement analysis. The DHM is used to predict the skin shape of a person when changing the posture through the linear blend skinning method combined with a data-driven approach to improve the accuracy. A personalized anatomically parametrized skeletal model is used to estimate the positions of joint centers and drive the skin deformations. The DHM parameters were determined using a database of 4D scans from adults (15 men and 15 women) performing sportive movements. Previous studies related with the DHM development are available.
At the current stage, the skin prediction quality is evaluated. To this end, the DHM was used to predict the posed skin shape from an initial shape in A-pose and the predictions were compared with scanned posed data. The posed skin predictions were performed for 4 women and 4 men with body mass indices (BMI) on the underweight, healthy weight, overweight, and obese categories. The predictions were performed in 6 varied sportive poses. The prediction quality was expressed as the average vertex-to-surface error for the body without head, hands and feet.
It was concluded that the DHM developed allows predicting the skin shape of people in the tested sportive postures starting from their A-pose shape. The evaluation was conducted on both men and women with a variety of body mass indexes (BMIs), from underweight to obese. This demonstrates the model's potential for broad applicability across different body shapes. The use of an anatomically parametrized skeletal model is a significant strength. This suggests the model's deformations are not just statistically inferred but are also constrained by a plausible underlying anatomy, which is crucial for biomechanical accuracy. In the following stages of the study it would be of interest to evaluate the skeletal model and its potential for movement analysis.