# US 2025/0292385 A1 · Image inspection apparatus, method for image inspection, and storage medium

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

- Number: US 2025/0292385 A1
- Document ID: US20250292385A1
- Status: Published application
- Kind: A1
- Application number: 19/070,610
- Filed: 2025-03-05
- Published: 2025-09-18
- Earliest claimed: 2024-03-15
- Assignees: Canon Kabushiki Kaisha
- Applicants: Canon Kabushiki Kaisha
- Inventors: Kenji KURIBAYASHI
- Web page: https://patentgrep.com/patent/US20250292385A1

## CPC

- G06T 7/0004: {Industrial image inspection}
- H04N 1/00076: {locally}
- G06T 2207/20081: Training; Learning
- G06T 2207/30144: Printing quality
- G06T 2207/30168: Image quality inspection
- H04N 1/00005: {relating to image data}
- H04N 1/00045: {using a reference pattern designed for the purpose, e.g. a test chart}
- H04N 1/00061: {using a remote apparatus (H04N1/00066 takes precedence)}
- H04N 1/00079: {remotely}
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

An image inspection apparatus that acquires information on a corrective action performed by a user for an image anomaly, and performs training of a learning model to estimate a candidate for corrective action content from an image anomaly using the image anomaly and the acquired information on the corrective action as learning data.

## Claims

### Claim 1

An image inspection apparatus, comprising:
an acquiring unit configured to acquire information on a corrective action performed by a user for an image anomaly; and
a training unit configured to perform training of a learning model for estimating a candidate of corrective action content from an image anomaly using, as learning data, the image anomaly and the acquired information on the corrective action.

### Claim 2 (depends on 1)

The image inspection apparatus according to claim 1, further comprising
an estimation unit configured to estimate a candidate of corrective action content for an image anomaly by the learning model,
wherein the acquiring unit acquires information on a corrective action performed by the user from the estimated candidate of the corrective action content.

### Claim 3 (depends on 1)

The image inspection apparatus according to claim 1, wherein the candidate of the corrective action content includes information on a faulty part.

### Claim 4 (depends on 2)

The image inspection apparatus according to claim 2,
wherein the candidate of the corrective action content includes information on a faulty part and a confidence of the corrective action content.

### Claim 5 (depends on 1)

The image inspection apparatus according to claim 1, further comprising
a transmitting unit configured to transmit, to an input device, the estimated candidate of the corrective action content,
wherein the acquiring unit acquires information on a corrective action performed by the user from the input device.

### Claim 6 (depends on 1)

The image inspection apparatus according to claim 1, further comprising
a determining unit configured to determine whether or not information on a replacement part included in the acquired corrective action exists in actually replaced part information,
wherein if the determining unit determines that the information does not exist, the information is not used as the learning data.

### Claim 7

A method for image inspection, the method comprising:
acquiring information on a corrective action performed by a user for an image anomaly; and
performing training of a learning model for estimating a candidate of corrective action content from an image anomaly using, as learning data, the image anomaly and the acquired information on the corrective action.

### Claim 8

A non-transitory computer-readable storage medium, the storage medium storing a program for causing a computer to execute each step of a method of controlling an image inspection apparatus, the method comprising:
acquiring information on a corrective action performed by a user for an image anomaly; and
performing training of a learning model for estimating a candidate of corrective action content from an image anomaly using, as learning data, the image anomaly and the acquired information on the corrective action.
