# US 2023/0067730 A1 · Information processing apparatus and control method thereof, and learning apparatus

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

- Number: US 2023/0067730 A1
- Document ID: US20230067730A1
- Status: Published application
- Kind: A1
- Application number: 17/878,354
- Filed: 2022-08-01
- Published: 2023-03-02
- Earliest claimed: 2021-08-31
- Assignees: Canon Kabushiki Kaisha
- Applicants: Canon Kabushiki Kaisha
- Inventors: Koichiro Nakamura
- Web page: https://patentgrep.com/patent/US20230067730A1

## CPC

- G06V 10/82: using neural networks
- G06V 10/764: using classification, e.g. of video objects
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 40/172: {Classification, e.g. identification}
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

An information processing apparatus comprises: a first obtaining unit configured to obtain feature vectors related to a first category from a vector group; a second obtaining unit configured to obtain feature vectors related to a second category from the vector group; a calculation unit configured to calculate a transformation matrix for converting a projection source vector obtained based on a feature vector obtained by the second obtaining unit into a projection destination vector obtained based on a feature vector obtained by the first obtaining unit; and a generation unit configured to generate a pseudo feature vector related to a third category by performing conversion processing on feature vectors included in the feature vectors obtained by the second obtaining unit using the transformation matrix.

## Claims

### Claim 1

An information processing apparatus comprising:
a processor; and
a memory containing instructions that, when executed by the processor, cause the processor to function as:
a first obtaining unit configured to obtain one or more feature vectors related to a first category, from a vector group including a plurality of feature vectors, each associated with a category;
a second obtaining unit configured to obtain from the vector group one or more feature vectors, at least one related to a second category different from the first category;
a calculation unit configured to calculate a transformation matrix for converting a projection source vector obtained based on a feature vector obtained by the second obtaining unit into a projection destination vector obtained based on a feature vector obtained by the first obtaining unit; and
a generation unit configured to generate a pseudo feature vector related to a third category different from the first category and the second category by performing conversion processing on one or more feature vectors included in the feature vectors obtained by the second obtaining unit using the transformation matrix.

### Claim 2 (depends on 1)

The information processing apparatus according to claim 1, further comprising:
a vector generation unit configured to generate a feature vector representing a feature of an image from the image; and
a holding unit configured to hold a plurality of feature vectors generated based on a plurality of images by the vector generation unit as the vector group.

### Claim 3 (depends on 2)

The information processing apparatus according to claim 2, further comprising:
an updating unit configured to update a generation parameter to be used for generation of a feature vector by the vector generation unit, wherein
the updating unit updates the generation parameter so as to reduce an inter-vector distance between feature vectors belonging to the same category, increase an inter-vector distance between feature vectors belong to different categories, and increase an inter-vector distance between an arbitrary feature vector included in the vector group and the pseudo feature vector.

### Claim 4 (depends on 3)

The information processing apparatus according to claim 3, wherein
the vector generation unit generates a feature vector from an image using a neural network, and
the generation parameter is a weighting coefficient of the neural network.

### Claim 5 (depends on 1)

The information processing apparatus according to claim 1, wherein
the first calculation means calculates the projection destination vector based on an average of the feature vectors obtained by the first obtaining unit, and
the second calculation means calculates the projection source vector based on an average of the feature vectors obtained by the second obtaining unit.

### Claim 6 (depends on 1)

The information processing apparatus according to claim 1, further comprising:
a second holding unit configured to hold as a second vector group a plurality of representative vectors, each a feature vector associated with a category and representing a related category, wherein
the first obtaining unit and the second obtaining unit obtain a representative vector from the second vector group, and
the generation unit generates a pseudo representative vector representing the third category.

### Claim 7 (depends on 6)

The information processing apparatus according to claim 6, further comprising:
a second updating unit configured to update the second vector group held by the second holding unit based on the pseudo representative vector generated by the generation unit.

### Claim 8 (depends on 6)

The information processing apparatus according to claim 6, further comprising:
a third updating unit configured to update the pseudo representative vector generated by the generation unit, wherein
the third updating unit updates so as to reduce an inter-vector distance between a neighbor pseudo representative vector within a predetermined inter-vector distance from an arbitrary feature vector included in the vector group and the arbitrary feature vector.

### Claim 9 (depends on 1)

A learning apparatus, comprising:
the information processing apparatus according to claim 1; and
a learning unit configured to perform learning using at least the pseudo feature vector generated by the information processing apparatus.

### Claim 10

A method for controlling an information processing apparatus, the method comprising:
first obtaining in which one or more feature vectors related to a first category are obtained from a vector group including a plurality of feature vectors, each related to a category;
second obtaining in which one or more feature vectors, at least one related to a second category different from the first category, are obtained from the vector group
calculating a transformation matrix for converting a projection source vector obtained based on a feature vector obtained in the second obtaining into a projection destination vector obtained based on a feature vector obtained in the first obtaining; and
generating a pseudo feature vector related to a third category different from the first category and the second category by performing conversion processing on one or more feature vectors included in the feature vectors obtained in the second obtaining using the transformation matrix.

### Claim 11

A non-transitory computer-readable recording medium storing a program for causing a computer to execute a method for controlling an information processing apparatus, the method comprising:
first obtaining in which one or more feature vectors related to a first category are obtained from a vector group including a plurality of feature vectors, each related to a category:
second obtaining in which one or more feature vectors, at least one related to a second category different from the first category, are obtained from the vector group
calculating a transformation matrix for converting a projection source vector obtained based on a feature vector obtained in the second obtaining into a projection destination vector obtained based on a feature vector obtained in the first obtaining; and
generating a pseudo feature vector related to a third category different from the first category and the second category by performing conversion processing on one or more feature vectors included in the feature vectors obtained in the second obtaining using the transformation matrix.
