# US 2021/0232866 A1 · Concurrent image and corresponding multi-channel auxiliary data generation for a generative model

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

- Number: US 2021/0232866 A1
- Document ID: US20210232866A1
- Status: Published application
- Kind: A1
- Application number: 17/230,366
- Filed: 2021-04-14
- Published: 2021-07-29
- Earliest claimed: 2019-03-28
- Assignees: General Electric Company
- Applicants: General Electric Company
- Inventors: Ravi Soni; Gopal B. Avinash; Min Zhang
- Web page: https://patentgrep.com/patent/US20210232866A1

## CPC

- G06V 10/454: {Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]}
- G06F 18/217: {Validation; Performance evaluation; Active pattern learning techniques}
- G06F 18/2415: based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F 18/2431: Multiple classes
- G06N 20/00: Machine learning
- G06V 10/764: using classification, e.g. of video objects
- G06V 10/82: using neural networks
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Systems and techniques for providing concurrent image and corresponding multi-channel auxiliary data generation for a generative model are presented. In one example, a system generates synthetic multi-channel data associated with a synthetic version of imaging data. The system also predicts multi-channel imaging data and the synthetic multi-channel data with a first predicted class set or a second predicted class set. Furthermore, the system employs the first predicted class set or the second predicted class set for the synthetic multi-channel data to train a generative adversarial network model.

## Claims

### Claim 1

A system, comprising:
a memory that stores computer executable components; and
a processor that executes the computer executable components stored in the memory, wherein the computer executable components comprise:
a multi-channel generator component that generates:
synthetic multi-channel data associated with a synthetic image,
a first data channel associated with a synthetic image,
a second data channel associated with first segmentation data indicative of a segmentation for the synthetic image, and
a third data channel associated with second segmentation data indicative of a remaining segmentation for the synthetic image; and
a post-processing component that generates first mask data indicative of a first ground truth mask for the first segmentation data, and second mask data indicative of a second ground truth mask for the second segmentation data.

### Claim 2 (depends on 1)

The system of claim 1, wherein the computer executable components further comprise:
a synthetic image generator component that generates the synthetic image.

### Claim 3 (depends on 1)

The system of claim 1, wherein the computer executable components further comprise:
a segmentation component that generates the first segmentation data associated with the second data channel and the second segmentation data associated with the third data channel.

### Claim 4 (depends on 1)

The system of claim 1, wherein the multi-channel generator component generates the synthetic multi-channel data based on a data distribution.

### Claim 5 (depends on 4)

The system of claim 4, wherein the data distribution is associated with a random vector.

### Claim 6 (depends on 1)

The system of claim 1, wherein the multi-channel generator component generates the synthetic multi-channel data based on at least one latent random variable.

### Claim 7 (depends on 1)

The system of claim 1, wherein the multi-channel generator component employs a deep neural network to generate the synthetic multi-channel data.

### Claim 8 (depends on 1)

The system of claim 1, wherein the first mask data comprises a binary mask that corresponds to an annotation associated with the first segmentation data indicative of the segmentation for the synthetic image.

### Claim 9 (depends on 8)

The system of claim 8, wherein the second mask data comprises a binary mask that corresponds to an annotation associated with the first segmentation data indicative of the remaining segmentation for the synthetic image.

### Claim 10 (depends on 1)

The system of claim 1, wherein the first mask data comprises a filter to mask one or more regions in the synthetic image.

### Claim 11 (depends on 1)

The system of claim 1, wherein the first mask data comprises a binary weight for segmentation in an image associated with the synthetic image.

### Claim 12 (depends on 1)

The system of claim 1, wherein the first mask data comprises a group of pixels that define a location for segmentation associated with an image, wherein the image is associated with the synthetic image.

### Claim 13

A method, comprising:
generating, by a system comprising a processor, synthetic multi-channel data associated with a synthetic image, comprising:
generating, by the system, a first data channel associated with the synthetic image,
generating, by the system, a second data channel associated with first segmentation data indicative of a segmentation for the synthetic image, and
generating, by the system, a third data channel associated with second segmentation data indicative of a remaining segmentation for the synthetic image;
generating, by the system, first mask data indicative of a first ground truth mask for the first segmentation data; and
generating, by the system, second mask data indicative of a second ground truth mask for the second segmentation data.

### Claim 14 (depends on 13)

The method of claim 13, wherein generating the synthetic multi-channel data comprises generating the synthetic multi-channel data based on a data distribution.

### Claim 15 (depends on 14)

The method of claim 14, wherein the data distribution is associated with a random vector.

### Claim 16 (depends on 13)

The method of claim 13, wherein generating the synthetic multi-channel data comprises generating the synthetic multi-channel data based on at least one latent random variable.

### Claim 17 (depends on 13)

The method of claim 13, wherein generating the synthetic multi-channel data comprises using a deep neural network to generate the synthetic multi-channel data.

### Claim 18

A non-transitory machine-readable medium, comprising executable instructions that, when executed by a processor, facilitate performance of operations, comprising:
generating a first data channel associated with a synthetic image;
generating a second data channel associated with first segmentation data indicative of a segmentation for the synthetic image;
generating a third data channel associated with second segmentation data indicative of a remaining segmentation for the synthetic image;
generating first mask data indicative of a first ground truth mask for the first segmentation data; and
generating second mask data indicative of a second ground truth mask for the second segmentation data.

### Claim 19 (depends on 18)

The non-transitory machine-readable medium of claim 18, wherein the first mask data comprises a filter to mask one or more regions in the synthetic image.

### Claim 20 (depends on 18)

The non-transitory machine-readable medium of claim 18, the first mask data comprises a binary weight for segmentation in an image associated with the synthetic image.
