欢迎访问《图学学报》 分享到：
图学学报

图学学报

• 专论：第16届媒体智能与大数据计算会议（CIDE & DEA 2019 大连） • 上一篇    下一篇

面向不同体型特征的服装款式迁移方法

  

  1. (1. 华北电力大学控制与计算机工程学院，北京 102206； 2. 中国科学院计算技术研究所，北京 100190)
  • 出版日期:2019-10-31 发布日期:2019-11-06
  • 基金资助:
     “十三五”重点研发计划项目(2017YFC0804900)；国家自然科学基金项目(61532002，61702482)；国家自然科学基金青年基金项目(61300131， 61972397)

Transfer Method of Body Shape-Oriented Garment Style

  1. (1. School of Control and Computer Engineering, North China Electric Power University, Beijing 102206, China;  2. Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100190, China)
  • Online:2019-10-31 Published:2019-11-06

PDF (PC)

215

摘要：

大部分成衣的设计以标准比例的人模作为参照，而对于非标准体型的顾客，标码 服装的大小尺寸则很难与之进行有效匹配。基于此，提出了一种面向不同体型特征的服装款 式迁移方法。首先，对于一批不同款式的服装，通过物理模拟的方式穿着到标模及非标模人 体之上，形成标模试穿的服装实例及非标模试穿的服装实例；其次，使用仿射变换表示同款 服装在标模及非标模下服装实例间的变形映射，并借助主成分分析法求解服装变形，在保留 因体型特征导致的服装形变的基础上，剔除由服装款式信息引起的服装形变；最后，将服装 变形用于标模到非标模的服装款式迁移，并使用平均离散曲率衡量迁移前后服装款式的改变 程度。实验结果表明，迁移后的服装携带了标模服装的款式信息，并保留了非标模服装的体 型特征。

关键词: 虚拟试衣, 服装款式迁移, 服装形变, 仿射变换, 主成分分析

Abstract: Most garments are designed according to the standard scale of mannequin. However, for customers with non-standard body shape, it is hard to match them with garments of standard size. Based on the above background, we have proposed a body shape-oriented garment style transfer method. Firstly, for a number of garments of different styles, physically-based simulation method is used to wear the garments on standard human model and the non-standard human model. The garment instances fitting for standard body shapes and non-standard body shapes, are then constructed. Secondly, affine transformation is used to represent deformation between the same garment instances under the standard body shape and non-standard body shape, and the cloth deformation is computed using principal component analysis so that the cloth deformation caused by body shape can be retained and the cloth deformation caused by the garment style can be removed. Finally, cloth deformation is used to transfer the garment style from standard to non-standard model, and we use average discrete curvature to measure the degree of change in clothing style before and after garment transfer. The experiment results show that the transferred garments have the style information of the garment under standard body shape and the body characteristics of the garment under non-standard body shape.

Key words: virtual fitting, garment style transfer, cloth deformation, affine transformation, principal component analysis