Automated grading of clothing patterns of garment

Patent metadata

Field	Value
Publication number	US-11308707-B2
Application number	US-202016996490-A
Country	US
Kind code	B2
Filing date	Aug 18, 2020
Priority date	Aug 19, 2019
Publication date	Apr 19, 2022
Grant date	Apr 19, 2022

Abstract

Provided is an automatic grading method including calculating a first strain ratio between a three-dimensional (3D) source avatar and a 3D target avatar, determining a mapping relationship between 3D source garment draped over the source avatar and a body portion of the source avatar, converting the source garment into 3D target garment draped over the target avatar, based on the first strain ratio and the mapping relationship, and outputting a two-dimensional (2D) target pattern constituting the target garment.

First claim

What is claimed is:

1. An automatic grading method for a garment, comprising:
determining avatar strain ratios, each of the avatar strain ratios representing differences in one or more source avatar polygons that are part of a mesh representing one or more body portions of a three-dimensional (3D) source avatar and one or more target avatar polygons that are part of another mesh representing corresponding one or more body portions of a 3D target avatar, the one or more body portions of the 3D source avatar having different shapes compared to the corresponding one or more body portions of the 3D target avatar;
determining mapping relationships between portions of a 3D source garment and the one or more body portions of the 3D source avatar, the 3D source garment dimensioned for draping the 3D source avatar and comprising a plurality of adjoined two-dimensional (2D) source patterns; and
determining shapes of 2D target patterns of a 3D target garment dimensioned for draping the 3D target avatar by processing the avatar strain ratios and the mapping relationships, the processing comprising:
determining a second rate between a source strain and a target strain candidate of a target pattern candidate, wherein the target strain candidate represents a target strain between a reference polygon and a polygon of the target pattern candidate; and
determining locations of the target pattern polygons that restrains a difference between a first rate determined based on the avatar strain ratios and the second rate by an optimization algorithm.

2. The automatic grading method of claim 1 , wherein determining the shapes of the 2D target patterns includes:
determining the one or more source avatar polygons mapped to each of source garment polygons as defined by the mapping relationships, wherein the one or more source avatar polygons are part of the mesh representing the one or more body portions of the 3D source avatar, and wherein the source garment polygons form a source mesh representing the 3D source garment;
deforming each of the source garment polygons into each of target garment polygons that form a target mesh representing the 3D target garment by applying a garment transfer function to each of the source garment polygons, wherein the garment transfer function is a function of one or more transformation functions that transform the one or more source avatar polygons mapped to each of the source garment polygons into each of the one or more target avatar polygons;
determining pattern strains between source pattern polygons corresponding to the source garment polygons and target pattern polygons corresponding to the target garment polygons; and
generating the 2D target patterns by applying the determined pattern strains to the plurality of adjoined 2D source patterns.

3. The automatic grading method of claim 2 , wherein each of the source garment polygons is deformed into each of the target garment polygons according to statistics on the avatar strain ratios.

4. The automatic grading method of claim 1 , further comprises performing relaxing of the first rate determined based on the avatar strain ratios before determining the locations of the target pattern polygons.

5. The automatic grading method of claim 4 , wherein the locations of the target pattern polygons are determined by further restraining deviations of curvatures of outlines of the 2D target patterns relative to curvature of outlines of the plurality of adjoined 2D source patterns.

6. The automatic grading method of claim 5 , wherein the locations of the target pattern polygons are determined by further restraining deviations of ratios of sewing lines between adjoining ones of the 2D target patterns relative to ratios of sewing lines of adjoining ones of the plurality of adjoined 2D source patterns.

7. The automatic grading method of claim 1 , wherein the avatar strain ratios are determined as deformations of unit polygons forming a mesh representing the 3D target avatar and corresponding unit polygons forming another mesh representing the 3D source avatar.

8. The automatic grading method of claim 1 , wherein the avatar strain ratios are determined by:
determining a transformation matrix indicating a deformation between a first tetrahedron defined by vertices of the one or more source avatar polygons forming the mesh representing the one or more body portions of the 3D source avatar and a second tetrahedron defined by vertices of the one or more target avatar polygons forming the other mesh representing the one or more body portions of the 3D target avatar, the one or more target avatar polygons corresponding to the one or more source avatar polygons, and
deriving the avatar strain ratios from the transformation matrix.

9. The automatic grading method of claim 1 , wherein the determining the mapping relationships includes:
determining a mapping relationship between a first pattern among the plurality of adjoined 2D source patterns constituting the 3D source garment and the one or more body portions of the 3D source avatar.

10. The automatic grading method of claim 9 , wherein determining the mapping relationships includes:
determining, for each source avatar polygon, a subset of source garment polygons closest to each source avatar polygon as portions of the 3D source garment mapped to the one or more body portions of the 3D source avatar, and
registering the subset of source garment polygons as being mapped to each source avatar polygon in a mapping list.

11. A non-transitory computer readable storage medium storing instructions thereon, the instructions when executed by a processor cause the processor to:
determine avatar strain ratios, each of the avatar strain ratios representing differences in one or more source avatar polygons that are part of a mesh representing one or more body portions of a three-dimensional (3D) source avatar and one or more target avatar polygons that are part of another mesh representing corresponding one or more body portions of a 3D target avatar, the one or more body portions of the 3D source avatar having different shapes compared to the corresponding one or more body portions of the 3D target avatar;
determine mapping relationships between portions of a 3D source garment and the one or more body portions of the 3D source avatar, the 3D source garment dimensioned for draping the 3D source avatar and comprising a plurality of adjoined two-dimensional (2D) source patterns; and
determine shapes of 2D target patterns of a 3D target garment dimensioned for draping the 3D target avatar by processing the avatar strain ratios and the mapping relationships, the processing comprising:
determining a second rate between a source strain and a target strain candidate of a target pattern candidate, wherein the target strain candidate represents a target strain between a reference polygon and a polygon of the target pattern candidate; and
determining locations of the target pattern polygons that restrains a difference between a first rate determined based on the avatar strain ratios and the second rate by an optimization algorithm.

12. The non-transitory computer readable storage medium of claim 11 , wherein the instructions causing the processor to determine the shapes of the 2D target patterns includes instructions to:
determine the one or more source avatar polygons mapped to each of source garment polygons as defined by the mapping relationships, wherein the one or more source avatar polygons are part of the mesh representing the one or more body portions of the 3D source avatar, and wherein the source garment polygons form a source mesh representing the 3D source garment;
deform each of the

Assignees

• Clo Virtual Fashion Inc

Inventors

• Oh Seungwoo

Classifications

• G06T17/20

  Finite element generation, e.g. wire-frame surface description, {tesselation} · CPC title

• G06T19/20 Primary

  Editing of three-dimensional [3D] images, e.g. changing shapes or colours, aligning objects or positioning parts · CPC title

• G06T2219/2021

  Shape modification · CPC title

• G06Q30/0643 Primary

  graphically representing goods, e.g. 3D product representation · CPC title

• G06T2210/16

  Cloth · CPC title