Page - 172 - in Joint Austrian Computer Vision and Robotics Workshop 2020
Image of the Page - 172 -
Text of the Page - 172 -
AUTOMATEDGENERATIONOF3DGARMENTSINDIFFERENTSIZES
FROMASINGLESCAN
StefanHauswiesner,PhilippGrasmug
ReactiveReality
{hauswiesner,grasmug}@reactivereality.com
Figure1: Resultsof themethod.
Abstract. We describe a method to generate addi-
tional sizes of a garment from a single scanned size
and grading tables. The method helps retailers and
manufacturers toefficientlycapturetheirentireprod-
uct range,which in turnenablesadvancedARappli-
cations suchasvirtual fashion try-on.
1. Introduction
Online fashion retailers need 3D models of their
entireproduct range toenableadvancede-commerce
applications, such as 3D viewing and virtual try-on.
These retailers usually have a high number of items
andtheproduct rangechangesfrequently. Therefore,
to obtain 3D models for their entire product catalog,
manualmodeling isnot a feasibleapproach.
3D reconstruction through photogrammetry is
more efficient and photo-realistic. However, retail-
ers want to avoid the overhead of scanning multiple
sizes of a single garment. This document describes
adifferentapproachwhichalgorithmicallygenerates
the different sizes from a single 3D reconstructed
model and the garment’s grading tables, and thereby
increases the scalability of photogrammetry. Figure
1shows results. (a) (b)
Figure 2: Sizing the mesh. a) semantic classification
map. Red parts do not scale with size, green parts
scale along a single dimension. b) sizing an upper
arm. The color coding shows the body part associa-
tion.
2.RelatedWork
Previous work in garment modeling generates
sizes by adapting a garment to a target body, as op-
posed to sizing tables [1]. Other machine-learning-
based approaches enable decomposition and assem-
bly of new garments but do not allow resizing [2] or
rely on templates [4], which inherently limits these
approaches toknownshapes.
3.Method
Thechallengeofsizesynthesis is thatgarmentsdo
not scale uniformly. For example, going from size
“Medium” to size “Large”, the scale factor for the
length of the sleeves is different from the factor for
thecircumferenceof thesleeve. Thewayagarment’s
parts scale is described by a grading table. We use
this information to adjust the geometry of the model
for thedistinct sizes.
Furthermore, the fabric of the garment is not
172
Joint Austrian Computer Vision and Robotics Workshop 2020
- Title
- Joint Austrian Computer Vision and Robotics Workshop 2020
- Editor
- Graz University of Technology
- Location
- Graz
- Date
- 2020
- Language
- English
- License
- CC BY 4.0
- ISBN
- 978-3-85125-752-6
- Size
- 21.0 x 29.7 cm
- Pages
- 188
- Categories
- Informatik
- Technik