# US 2021/0279851 A1 · Manufacturing method of learning data, learning method, learning data manufacturing apparatus, learning apparatus, and memory medium

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

- Number: US 2021/0279851 A1
- Document ID: US20210279851A1
- Status: Published application
- Kind: A1
- Application number: 17/193,429
- Filed: 2021-03-05
- Published: 2021-09-09
- Earliest claimed: 2020-03-09
- Assignees: Canon Kabushiki Kaisha
- Applicants: Canon Kabushiki Kaisha
- Inventors: Takashi Oniki
- Web page: https://patentgrep.com/patent/US20210279851A1

## CPC

- G06T 5/73: Deblurring; Sharpening
- G06T 7/0002: {Inspection of images, e.g. flaw detection}
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/09: Supervised learning
- G06T 5/60: using machine learning, e.g. neural networks
- G06T 11/00: Two-dimensional [2D] image generation
- G06N 3/048: Activation functions
- G06T 2207/10024: Color image
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/30168: Image quality inspection
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A manufacturing method of learning data is used for making a neural network perform learning. The manufacturing method of learning data includes a first acquiring step configured to acquire an original image, a second acquiring step configured to acquire a first image as a training image generated by adding blur to the original image, and a third acquiring step configured to acquire a second image as a ground truth image generated by adding blur to the original image. A blur amount added to the second image is smaller than that added to the first image.

## Claims

### Claim 1

A manufacturing method of learning data used for making a neural network perform learning, the manufacturing method of learning data comprising:
a first acquiring step configured to acquire an original image;
a second acquiring step configured to acquire a first image as a training image generated by adding blur to the original image; and
a third acquiring step configured to acquire a second image as a ground truth image generated by adding blur to the original image,
wherein a blur amount added to the second image is smaller than that added to the first image.

### Claim 2 (depends on 1)

The manufacturing method of learning data according to claim 1,
wherein the blur is added based on optical information in each of the second acquiring step and the third acquiring step.

### Claim 3 (depends on 2)

The manufacturing method of learning data according to claim 2,
wherein the blur amount added to the second image is adjusted based on the optical information.

### Claim 4 (depends on 1)

The manufacturing method of learning data according to claim 1, further comprising a fourth acquiring step configured to acquire optical information,
wherein the optical information is a plurality of pieces of optical information each of which is different depending on an image height,
wherein the second acquiring step acquires the first image generated by adding blur using the plurality of pieces of optical information, and
wherein the third acquiring step acquires the second image generated by adding blur using the plurality of pieces of optical information.

### Claim 5 (depends on 1)

The manufacturing method of learning data according to claim 1,
wherein the second image is generated by adding blur to the original image, the blur having a different amount depending on an image height.

### Claim 6 (depends on 5)

The manufacturing method of learning data according to claim 5,
wherein the second image is generated by adding blur with a first blur amount to a first image height of the original image, and by adding blur with a second blur amount larger than the first blur amount to a second image height higher than the first image height of the original image.

### Claim 7 (depends on 1)

The manufacturing method of learning data according to claim 1, further comprising a fourth acquiring step configured to acquire optical information,
wherein the optical information is a plurality of pieces of optical information each of which is different depending on an image height,
wherein the second acquiring step acquires the first image which is generated by adding blur using optical information corresponding to an image height corresponding to the first image among the plurality of pieces of optical information, and
wherein the third acquiring step acquires the second image which is generated by adding blur using optical information corresponding to an image height corresponding to the second image among the plurality of pieces of optical information.

### Claim 8 (depends on 1)

The manufacturing method of learning data according to claim 1,
wherein in the third acquiring step, blur to be added is adjusted based on a luminance value.

### Claim 9 (depends on 8)

The manufacturing method of learning data according to claim 8,
wherein when the luminance value is a first luminance value, the second image is generated by adding blur with a first blur amount to the original image, and
wherein when the luminance value is a second luminance value lower than the first luminance value, the second image is generated by adding blur with a second blur amount larger than the first blur amount to the original image.

### Claim 10 (depends on 1)

A learning method that makes a neural network perform learning using learning data generated by the manufacturing method of learning data according to claim 1, the learning method comprising steps of:
generating a processed image by inputting the first image as the training image into the neural network; and
updating the neural network based on the second image and the processed image.

### Claim 11

A learning data manufacturing apparatus which manufactures learning data used for learning of a neural network, the learning data manufacturing apparatus comprising at least one processor or circuit configured to execute a plurality of tasks including:
a first acquisition task configured to acquire an original image;
a second acquisition task configured to acquire a first image generated by adding blur to the original image; and
a third acquisition task configured to acquire a second image generated by adding blur to the original image,
wherein a blur amount added to the second image is smaller than that added to the first image.

### Claim 12 (depends on 11)

A learning apparatus comprising:
the learning data manufacturing apparatus according to claim 11; and
at least one processor or circuit configured to execute a plurality of tasks including:
a generation task configured to generate a processed image by inputting the first image as the training image to a neural network; and
an update task configured to update the neural network based on the second image and the processed image.

### Claim 13 (depends on 1)

A non-transitory computer-readable storage medium storing a computer program that causes a computer to execute the manufacturing method of learning data according to claim 1.

### Claim 14 (depends on 1)

A manufacturing method of learnt model that manufactures a learnt model by making a neural network perform learning using learning data generated by the manufacturing method of learning data according to claim 1, the manufacturing method of learnt model comprising steps of:
generating a processed image by inputting the first image as the training image into the neural network; and
making the neural network perform learning based on the second image and the processed image.
