# US 2023/0147515 A1 · Boundary Correspondence Determination for Digital Objects

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0147515 A1
- Document ID: US20230147515A1
- Status: Published application
- Kind: A1
- Application number: 17/521,313
- Filed: 2021-11-08
- Published: 2023-05-11
- Earliest claimed: 2021-11-08
- Assignees: Adobe Inc.
- Applicants: Adobe Inc.
- Inventors: Souymodip Chakraborty; Vineet Batra; Ankit Phogat
- Web page: https://patentgrep.com/patent/US20230147515A1

## CPC

- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06T 11/23: using straight lines or curves
- G06N 3/045: Combinations of networks
- G06N 3/08: Learning methods
- G06T 7/344: {involving models}
- G06T 11/40: Filling planar surfaces by adding surface attributes, e.g. adding colours or textures
- G06T 13/80: Two-dimensional [2D] animation, e.g. using sprites
- G06N 3/048: Activation functions
- G06N 3/09: Supervised learning
- G06T 2207/10016: Video; Image sequence
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20164: Salient point detection; Corner detection
- G06T 2207/30196: Human being; Person
- G06T 2207/30221: Sports video; Sports image
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

Boundary correspondence determination techniques are described for digital objects as implemented by a boundary correspondence system. In an implementation, the boundary correspondence system partitions outer boundaries of first and second digital objects into a first plurality of cuts and a second plurality of cuts, respectively. A set of corresponding cut pairs are then determined based on a comparison of the first plurality of cuts with the second plurality of cuts. Further, corresponding anchor point pairs are determined based on a comparison of anchor points of the set of corresponding cut pairs. The boundary correspondence system then generates and outputs a mapping of the first digital object to the second digital object based on the determined correspondence of anchor point pairs.

## Claims

### Claim 1

In a digital medium digital object correspondence determination environment, a method implemented by a computing device, the method comprising:
partitioning, by the computing device, outer boundaries of first and second digital objects into a first plurality of cuts and a second plurality of cuts, respectively;
comparing, by the computing device, the first plurality of cuts with the second plurality of cuts;
determining, by the computing device, a set of corresponding cut pairs based on the comparing, each said cut pair having a cut from the first plurality of cuts and a cut from the second plurality of cuts;
determining, by the computing device, correspondence of anchor point pairs, one to another, by comparing anchor points of the set of corresponding cut pairs;
generating, by the computing device, a mapping of the first digital object to the second digital object based on the determined correspondence of the anchor point pairs; and
outputting, by the computing device, the mapping.

### Claim 2 (depends on 1)

The method as described in claim 1, wherein the first plurality of cuts includes multiple cuts of different lengths each centered at a first respective point along the outer boundary of the first digital object, and the second plurality of cuts includes multiple cuts of different lengths each centered at a second respective point along the outer boundary of the second digital object.

### Claim 3 (depends on 1)

The method as described in claim 1, wherein comparing the first plurality of cuts with the second plurality of cuts includes determining a measure of similarity for each respective cut from the first plurality of cuts with each respective cut from the second plurality of cuts.

### Claim 4 (depends on 1)

The method as described in claim 1, wherein comparing the first plurality of cuts with the second plurality of cuts includes determining and aligning principal axes of the first plurality of cuts and the second plurality of cuts.

### Claim 5 (depends on 1)

The method as described in claim 1, wherein comparing the first plurality of cuts with the second plurality of cuts includes leveraging a deep neural network trained on a list of tuples, each said tuple containing two images and an indication of whether the two images are similar or dissimilar.

### Claim 6 (depends on 5)

The method as described in claim 5, wherein a tuple of the list of tuples is generated by:
partitioning, by the computing device, at least one training boundary into a plurality of seed training cuts;
selecting, by the computing device, control points along a seed training cut of the plurality of seed training cuts, the seed training cut corresponding to a first image of the two images; and
generating, by the computing device, an altered training cut by arbitrarily moving the control points of the seed training cut by a perturbation factor, the altered training cut corresponding to a second image of the two images, and the two images indicated as similar or dissimilar based on the perturbation factor.

### Claim 7 (depends on 1)

The method as described in claim 1, wherein the set of corresponding cut pairs are determined, in part, by identifying a first cut pair based on similarity of the first plurality of cuts with the second plurality of cuts.

### Claim 8 (depends on 7)

The method as described in claim 7, wherein additional cut pairs in the set are determined by iteratively processing cuts of the first and second pluralities of cuts adjacent to the first cut pair.

### Claim 9 (depends on 1)

The method as described in claim 1, wherein the correspondence of the anchor point pairs is determined based on similarity of portions of the outer boundaries of the first and second digital objects surrounding the anchor point pairs and an amount of deformation used to match positioning of the anchor point pairs.

### Claim 10 (depends on 1)

The method as described in claim 1, wherein the mapping is represented by an ordered sequence of the anchor point pairs disposed along the outer boundaries of the first and second digital objects, each said anchor point pair being associated with a same value in the ordered sequence.

### Claim 11 (depends on 1)

The method as described in claim 1, wherein the outer boundaries of the first and second digital objects are sets of Bezier segments.

### Claim 12 (depends on 1)

The method as described in claim 1, further comprising:
receiving the first and second digital objects as Raster images; and
defining the outer boundaries of the first and second digital objects based on outlines of the Raster images.

### Claim 13

In a digital medium digital object correspondence determination environment, a system comprising:
a boundary partitioning module implemented at least partially in hardware of a computing device to partition outer boundaries of first and second digital objects into a first plurality of cuts and a second plurality of cuts, respectively;
a cut pairing module implemented at least partially in hardware of the computing device to determine a set of corresponding cut pairs based on a comparison of the first plurality of cuts with the second plurality of cuts, each said cut pair having a cut from the first plurality of cuts and a cut from the second plurality of cuts; and
a mapping module implemented at least partially in hardware of the computing device to generate and output a mapping of the first digital object to the second digital object based on a correspondence of anchor point pairs determined from a comparison of anchor points of the corresponding cut pairs in the set.

### Claim 14 (depends on 13)

The system as described in claim 13, wherein the first plurality of cuts includes multiple cuts of different lengths each centered at a first respective point along the outer boundary of the first digital object, and the second plurality of cuts includes multiple cuts of different lengths each centered at a second respective point along the outer boundary of the second digital object.

### Claim 15 (depends on 13)

The system as described in claim 13, wherein the cut pairing module is configured to leverage a deep neural network trained on a list of tuples to determine a measure of similarity for each respective cut from the first plurality of cuts with each respective cut from the second plurality of cuts.

### Claim 16 (depends on 15)

The system as described in claim 15, further comprising:
a training module implemented at least partially in hardware of the computing device to generate the list of tuples, each said tuple including:
a seed training cut partitioned from an input boundary;
an altered training cut generated by arbitrary movement of control points disposed on the seed training cut by a perturbation factor; and
an indication of whether the seed training cut and the altered training cut are similar or dissimilar based on the perturbation factor.

### Claim 17

In a digital medium digital object correspondence determination environment, a system comprising:
a processing system; and
a computer-readable storage medium having instructions stored thereon that, responsive to execution by the processing system, causes the processing system to perform operations including:
partitioning outer boundaries of first and second digital objects into a first plurality of cuts and a second plurality of cuts, respectively;
determining a set of corresponding cut pairs by comparing the first plurality of cuts with the second plurality of cuts, each said cut pair having a cut from the first plurality of cuts and a cut from the second plurality of cuts;
determining correspondence of anchor point pairs, one to another, by comparing anchor points of the set of corresponding cut pairs;
generating a mapping of the first digital object to the second digital object based on the determined correspondence of the anchor point pairs; and
outputting the mapping.

### Claim 18 (depends on 17)

The system as described in claim 17, wherein the set of corresponding cut pairs include a first cut pair determined based on similarity of the first plurality of cuts with the second plurality of cuts and additional cut pairs determined by iteratively processing cuts of the first and second pluralities of cuts adjacent to the first cut pair.

### Claim 19 (depends on 17)

The system as described in claim 17, wherein the correspondence of the anchor point pairs is determined based on similarity of portions of the outer boundaries of the first and second digital objects surrounding the anchor point pairs and an amount of deformation used to match positioning of the anchor point pairs.

### Claim 20 (depends on 17)

The system as described in claim 17, wherein the mapping is represented by an ordered sequence of anchor point pairs disposed along the outer boundaries of the first and second digital objects, each said anchor point pair being associated with a same value in the ordered sequence.
