Design and Product Development

Automatic Human Body Measurement for Virtual Fitting Using Deep Learning: The Scan Avatar-Captured 2D Image Dataset

Authors
  • Seowoo Kim (friday513@snu.ac.kr)
  • Hyunwook Kim
  • Yesung Cho orcid logo (Seoul Nationl University)
  • Sohyun Lee
  • Juyeon Park (Seoul National University)

Abstract

The interest in contact-free body measurement using 2D images is growing since they can be applied to functions for creating avatars and checking how well clothes fit the consumers and be actively used in virtual fitting platforms. This study aims to use the deep learning method for virtual fitting by using the circumference of the human body as the feature value for the automatic measurement of the human body. This study further attempts to present a new method for human body measurement by using 3D body scan data to create learning data for deep learning. The developed algorithm have been improved and that the measurement of the human body using deep learning is significant. Also, the feature value by dividing the human body equally, not by selecting the landmarks and utilizes deep learning instead of calculating perimeter values of human body and building new training data by using the existing 3D human body DB (Size Korea) rather than the traditional method.

Keywords: human body measurements, 2D image, deep learning

How to Cite:

Kim, S., Kim, H., Cho, Y., Lee, S. & Park, J., (2022) “Automatic Human Body Measurement for Virtual Fitting Using Deep Learning: The Scan Avatar-Captured 2D Image Dataset”, International Textile and Apparel Association Annual Conference Proceedings 79(1). doi: https://doi.org/10.31274/itaa.16051

Downloads:
Download PDF
View PDF

293 Views

69 Downloads

Published on
31 Dec 2022
Peer Reviewed