3DBODY.TECH 2018 - Paper 18.131

S. Goonatilake and M. Ransley, "Applying Deep-Learning to Reconstruct Accurate 3D Body Maps Using Mobile Phones", in Proc. of 3DBODY.TECH 2018 - 9th Int. Conf. and Exh. on 3D Body Scanning and Processing Technologies, Lugano, Switzerland, 16-17 Oct. 2018, pp. 131.


Applying Deep-Learning to Reconstruct Accurate 3D Body Maps Using Mobile Phones



1 University College London, London, UK;
2 Bodymetrics, London, UK


In this paper, we present the application of Generative Adversarial Network (GAN) deeplearning neural networks to build a 3D body-model using a very large - 100,000 body-scan - proprietary database. We learn the relationships between multi-viewpoints and the corresponding body-measurements and also use a hierarchical PCA in the semantic segmentation of body-information. We then use recent Augmented Reality functionality of mobile phones (ARkit on ios) to determine a Height reference for a person and then ask the person to rotate with their arms slightly apart. The video of this action, sampled at 25 framesper-second, provides a very large data set to fit our learnt body-models and derive measurements and body-shape information. The approach also takes into account differences in body-measurements due to breathing. The results of several female and male models bodyscanned using this method were compared with output from a TC2 KX16 body-scanner. The results were comparable, within 1 cm deviation on key measurements - bust/chest, waist and hips. We believe that this opens up the possibility of large scale adoption of 3D body-mapping for millions of consumers worldwide.


Abstract: 18131goonatilake_abs.pdf
Proceedings: 3DBODY.TECH 2018, 16-17 Oct. 2018, Lugano, Switzerland
Pages: 131

Copyright notice

© Hometrica Consulting - Dr. Nicola D'Apuzzo, Switzerland, www.hometrica.ch.
Reproduction of the proceedings or any parts thereof (excluding short quotations for the use in the preparation of reviews and technical and scientific papers) may be made only after obtaining the specific approval of the publisher. The papers appearing in the proceedings reflect the author's opinions. Their inclusion in these publications does not necessary constitute endorsement by the editor or by the publisher. Authors retain all rights to individual papers.

Proceedings of 3DBODY.TECH International Conferences on 3D Body Scanning & Processing Technologies, © Hometrica Consulting, Switzerland