Recommending pattern designs for objects using a sequence-based machine-learning model

Patent metadata

Field	Value
Publication number	US-11093660-B2
Application number	US-201514968870-A
Country	US
Kind code	B2
Filing date	Dec 14, 2015
Priority date	Dec 14, 2015
Publication date	Aug 17, 2021
Grant date	Aug 17, 2021

Abstract

Methods and systems for aiding users in generating object pattern designs with increased speed. In particular, one or more embodiments train a sequence-based machine-learning model using training objects, each training object including a plurality of regions with a plurality of design elements. One or more embodiments identify a plurality regions of an object with a first region adjacent a second region. One or more embodiments receive a user selection of a design element for populating the first region with a first design element from a plurality of design elements. One or more embodiments identify a second design element from the plurality of design elements based on the first design element using the trained sequence-based machine-learning model. One or more embodiments also populate the second region with one or more instances of the second design element.

First claim

What is claimed is:

1. In a digital medium environment for generating an object pattern design, a method comprising:
identifying a plurality of circular regions of a turntable object based on visual characteristics indicating visible borders of the plurality of circular regions, the plurality of circular regions comprising a first region adjacent a second region arranged concentrically around a center of the turntable object;
receiving, via a graphical user interface, a user selection of a first design element from a plurality of design elements displayed within the graphical user interface;
determining, by at least one processor, a first composition of the first design element, the first composition defining composition attributes of the first design element;
populating, within the graphical user interface, the first region with a plurality of instances of the first design element based on the composition attributes of the first design element;
determining, by the at least one processor utilizing a hidden Markov model to analyze the first design element and the first region, a first hidden state value of the first region based on the first design element and a location or a size of the first region, wherein the hidden Markov model comprises a machine-learning model trained on a plurality of training objects including a plurality of training design elements;
generating, by the at least one processor utilizing the hidden Markov model, a second hidden state value for the second region based on the first hidden state value, an area of the second region, and a position of the second region relative to the first region;
identifying, by the at least one processor, a second design element from the plurality of design elements based on the second hidden state value;
determining, by the at least one processor, a second composition of the second design element, the second composition defining composition attributes of the second design element; and
populating, within the graphical user interface, the second region with a plurality of instances of the second design element based on the composition attributes of the second design element.

2. The method as recited in claim 1 , wherein determining the first composition of the first design element comprises determining a size value, a density value, and an arrangement value for the first design element.

3. The method as recited in claim 2 , wherein populating the first region with the plurality of instances of the first design element comprises populating the first region with the plurality of instances according to the size value, the density value, and the arrangement value for the first design element.

4. The method as recited in claim 1 , further comprising:
determining at least one characteristic of the first region;
determining the first composition of the first design element based on the at least one characteristic of the first region; and
populating the first region with the plurality of instances of the first design element based on the composition attributes of the first design element.

5. The method as recited in claim 1 , wherein determining the second composition of the second design element comprises determining the second composition based on the second hidden state value of the second region.

6. The method as recited in claim 1 , further comprising:
generating a third hidden state value of a third region from the plurality of circular regions based on the second hidden state value, the third region adjacent to the second region;
identifying a third design element from the plurality of design elements based on the third hidden state value; and
populating the third region with at least one instance of the third design element.

7. The method as recited in claim 1 , further comprising:
generating a third hidden state value of a third region from the plurality of circular regions based on the first hidden state value, the third region on a first side of the first region and the second region on a second side of the first region;
identifying a third design element from the plurality of design elements based on the third hidden state value; and
populating the third region with at least one instance of the third design element.

8. The method as recited in claim 1 , wherein generating the second hidden state value further comprises generating, utilizing the hidden Markov model, the second hidden state value based on a third hidden state value corresponding to a third region from the plurality of circular regions, the second region between the first region and the third region.

9. The method as recited in claim 1 , wherein identifying the second design element from the plurality of design elements based on the second hidden state value comprises:
calculating, utilizing the hidden Markov model, a style value based on the second hidden state value for the second region; and
selecting the second design element from the plurality of design elements based on the calculated style value.

10. In a digital medium environment for generating an object pattern design, a method comprising:
identifying a first object from images comprising a set of training objects, the object comprising a plurality of regions;
detecting, by at least one processor utilizing image processing to analyze the images comprising the set of training objects, a first design element within a first region of the plurality of regions and a second design element within a second region of the plurality of regions, the first region adjacent the second region;
generating, by the at least one processor, a first hidden state value for the first region based on the first design element and a location or a size of the first region and a second hidden state value for the second region based on the second design element and a location or a size of the second region, wherein the first hidden state value represents an overall design of the first region and the second hidden state value represents an overall design of the second region;
training, by the at least one processor, a hidden Markov model based on determining a relationship between the first hidden state value and the second hidden state value;
identifying a second object comprising a plurality of empty circular regions arranged concentrically around a center of the second object;
generating, utilizing the trained hidden Markov model, a plurality of hidden state values comprising a hidden state value for each of the plurality of empty circular regions based on locations and sizes of the plurality of empty circular regions;
determining, for the plurality of empty circular regions and based on the plurality of hidden state values, a plurality of design elements comprising a plurality of compositions defining composition attributes of the plurality of design elements for insertion into the plurality of empty circular regions; and
populating, within a graphical user interface and by the at least one processor, the plurality of empty circular regions in the second object with the plurality of design elements based on the plurality of compositions for the plurality of design elements.

11. The method as recited in claim 10 , further comprising:
identifying a plurality of training design elements from the set of training objects;
determining a style histogram for each design element from the plurality of training design elements;
comparing the style histograms for the plurality of training design elements; and
grouping the plurality of design elements into a plurality of clusters based on the style histograms for the plurality of training design elements.

12. The method as recited in claim 11 , wher

Assignees

• Adobe Inc

Inventors

• Asente Paul
• Lu Jingwan
• Phan Huy

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F30/00 Primary

  Computer-aided design [CAD] · CPC title

• G06F30/27 Primary

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G06N7/005

  Physics · mapped topic