3D Clothing: Garment Prepositioning & Human Body Adaptationimi.ntu.edu.sg/NewsEvents/Events/PastSeminars/Documents/... · 2019-03-06 · Garment Prepositioning • Preposition on

Zhang Yuzhe

SCE & IMI

Supervisor: Assoc Prof Zheng Jianmin

Co-Supervisor: Prof Nadia M. Thalmann

2011-11-22

3D Clothing:

Garment Prepositioning & Human Body Adaptation

Outline

• Introduction

• Garment Prepositioning

• Anthropometry based Body Adaptation

• Future work

• Q & A

Introduction

• Project – 3D clothing

• Motivation – Digital fashion applications

• Design, sew and simulate 2D patterns of garment on mannequin

• Try to make the whole process more intelligent and automatic

– Two main elements in digital fashion

• Human body model

• Patterns & Garment

Garment Prepositioning

• Problem:

– Input: 2D patterns, 3D body model

– Output: 3D garment mesh on body model

– Good prepositioning

• decrease computation of simulation

– Bad prepositioning

• lead to failure of simulation

Failure of Simulation


• Tasks involved

– Connectivity Task

• Pattern assembly

– Geometry Task

• 3D placement

• Triangulation

stitching

Placement


• State-of-the-art

– Manual approach

• Fashionizer

– Semi-automatic approach

• “Automatic pre-positioning of virtual clothing”, C. Grob, SCCG '03

• “ClothAssembler: a CAD module for feature-based garment pattern

assembler”, M. Fontana, J. Computers and Graphics, 06

• “From Interactive Positioning to Automatic Try-On of Garments”, T.

Thanh, MIRAGE '09

Body Segment

Flatten Bounding Surface

Arranging Patterns

Figurines Generic

Silhouette Real

Silhouette


• Our Approach

– Basic ideas

• Define our own format for connectivity

• Use template for reference

• Preposition-deformation for geometry

– Workflow

2D

patterns

Body

template

2D Editor

Own

format

3D

Body

Map to

template

Topology

stitching Preposition

3D

garment

Correspond


• Data Format

– Geometrical Information

• 2D Contour

• Mesh for 2D patterns

• Reference point and normal

– Topological Information

• Seam line

– Semantic Information

• Markers (link with skeleton)

• Other knowledge (Front/Back/Fold, Layer, Accessory )

Semantic info

Topological info

Geometrical info

Embedded geometrical & topological Info


• Map patterns to template body

– Preprocess

• Fold and turn over patterns if necessary

• Normalize patterns of garment to origin

– Semantic based matching

• Heuristic rules based on connectivity among patterns for basic

garment types

• Semantic knowledge embedded in template

Preprocess

Map onto template


• Correspond Input body to template

– Anthropometry extraction

– Skeleton and measurement matching

match


• Preposition on input body

– Deformation according to corresponding skeleton and measurement

– Minor transformation so that seams do not intersect with body

• Topology stitching

– Assembly patterns into the initial 3D mesh of garment by stitching

seams

Deformation Stitching


• Example (Own format Load body Preposition Simulation)

- Template body

- Customized body with different size & pose

Anthropometry based Body Adaptation

• Problem

– Model and customize an animated human body model

– Input: parameter, template body (with skeleton and skinning info)

– Output: adapted body

• Motivation

– Fast and easy to generate a group of human body model

– Anthropometry based (parameter linked with cloth measurement)

Parameter

control

Animate


• Tasks involved

– Anthropometric measurements modeling

• 24 measurements

– Girth: waist girth, hip girth

– Length: thigh length, arm length

– Non-geometric: weight

– Analysis of anthropometric measurements

• Limited input parameters

• Model the correlation among measurements

– Deformation under the control of measurements

– Animate model

• Skeleton & skinning information

• State-of-the-art

– Creative Approach

• “An example-based approach to human body manipulation “<2004>

• “Customizing & populating animated digital mannequins for real-

time application”<2010>

• …

– Disadvantages

• Treat the parameters independently

• Do not capture the detailed shape variations


• Our approach

– Basic ideas

• Predict full set of parameters from partial ones by analyzing training data

• Customize mesh and skeleton from full set anthropometric measurements

• Interactive modification by detail shape editing

– Workflow


Input 1 Full set of

anthropometric

measurements

Girth &

Length

Deformation

Detail Shape

Editing

Analyze

Deformation

Template

with full

setup

Customized

Animated

model

Animate Motion data

Partial

measure-

ments

Input 2


• Anthropometry analysis

– Predict the full set measurements w based on correlation

coefficient analysis

Partial

input

α

Training

data

Correlation

Coefficient

Analysis

wnormal

wcurrent Local

Modify wpersonal

Weighted

Combine Full set

w

Input Output

c

Pj

Pj+1 P’j

P’j+1


• Anthropometry based deformation

– Girth deformation

• Scale feature points on girth

• Laplacian reconstruction with feature constraints

Waist Girth=59.31cm Waist Girth=71.31cm


• Anthropometry based deformation

– Length deformation

• Linear scale skeleton

• Scale vertices in related regions

• Translate related girth measurement

• Propagate translation to its descendant nodes

∆l

Thigh Length=39.59cm Thigh Length=45.59cm


• Detail shape editing

– Set intersection plane, select target curve to be deformed

– Approximate target curve by B-Spline curve

– Deform curve by moving control points

– Deformation by Laplacian reconstruction

with constraints


• Experiment results

– Evaluation of anthropometry analysis

– Comparison with independent

linear scale method


• More example

– Body adaptation with limited anthropometric measurements

– Animation

• Used in garment simulation example

Future work

• Automatic patterns assembly and semantic representation

– Learning the semantic information based on shape analysis

– Automatic assembly patterns based on puzzle matching

• Automatic anthropometry extraction of input body

– Extract skeleton and measurements of individual body

– Customized template into individual body automatically according

to extracted knowledge

• Garment resize and flatten

– Resize 3D garment according to the measurements of body model

– Resize 2D patterns by 3D garment flatten

Thank you !

Q & A