IP Library › Granted Patent US 12,738,028
Granted Patent B2
US 12,738,028 · App. 18/255,170 · Granted Sep 15, 2026

Information processing apparatus, information processing method, imaging device, vehicle device, and medical robot device

Inventor: Kenji Suzuki (Tokyo, JP)
Assignee: SONY GROUP CORPORATION
G06V10/774A61B34/32G06V20/56A61B2034/2065B60W30/00B60W60/00B60W2420/403G06V10/764G06V10/82G06V2201/03
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,738,028
App. No.
18/255,170
Granted
Sep 15, 2026
Kind
B2
Abstract

There is provided an information processing apparatus for artificially increasing data of a minority attribute to generate learning data for making fair determination on each piece of input data. The information processing apparatus includes a data holding unit configured to hold first learning data to be used for learning a machine learning model an acquisition unit configured to acquire information regarding bias of the learning data a data generation unit configured to generate second learning data by using data included in the learning data on the basis of the information regarding bias and a learning unit configured to learn the machine learning model by using the first learning data and the second learning data.

Claims (69)

1 . An information processing apparatus, comprising:

circuitry configured to:

hold first learning data for a learning operation of a machine learning model;

acquire information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimpose noise on the first learning data;

generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and

execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model.

2 . The information processing apparatus according to claim 1 , wherein the circuitry is further configured to:

acquire information indicating a minority attribute in the held first learning data; and

generate, further based on the information indicating the minority attribute, the second learning data, wherein the generated second learning data have a same attribute as the minority attribute.

3 . The information processing apparatus according to claim 2 , wherein

the circuitry is further configured to generate, based on the minority attribute, Adversarial Example as the second learning data.

4 . The information processing apparatus according to claim 3 , wherein

the circuitry is further configured to generate the Adversarial Example further based on a Fast Gradient Sign Method.

5 . An information processing method, comprising:

storing first learning data for a learning operation of a machine learning model;

acquiring information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimposing noise on the first learning data;

generating second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and

executing, based on the generated second learning data and the stored first learning data, the learning operation of the machine learning model.

6 . A non-transitory computer-readable medium having stored thereon, computer-executable instructions which, when executed by a computer, cause the computer to execute operations, the operations comprising:

holding first learning data for a learning operation of a machine learning model;

acquiring information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimposing noise on the first learning data;

generating second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data; and

executing, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model.

7 . An imaging device, comprising:

circuitry configured to:

hold first learning data for a learning operation of a machine learning model;

acquire information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimpose noise on the first learning data;

generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;

execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;

capture an image; and

recognize the captured image based on the machine learning model and the executed learning operation of the machine learning model.

8 . The imaging device according to claim 7 , wherein the circuitry is further configured to:

determine, based on a field associated with the imaging device, a minority in the first learning data; and

generate the second learning data based on data of an image associated with the determined minority in the first learning data.

9 . A vehicle device, comprising:

circuitry configured to:

hold first learning data for a learning operation of a machine learning model;

acquire information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimpose noise on the first learning data;

generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;

execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;

capture an image in a vicinity of the vehicle device; and

recognize the captured image based on the machine learning model and

the executed learning operation of the machine learning model.

10 . A medical robot device, comprising:

circuitry configured to:

hold first learning data for a learning operation of a machine learning model;

acquire information regarding bias of the first learning data, wherein

the acquired information regarding the bias corresponds to a set of data associated with at least one specific attribute, and

the first learning data includes the set of data;

superimpose noise on the first learning data;

generate second learning data based on the superimposed noise on the first learning data and the information regarding the bias of the first learning data;

execute, based on the generated second learning data and the held first learning data, the learning operation of the machine learning model;

capture an image in a vicinity of a surgical site; and

recognize the captured image based on the machine learning model and the executed learning operation of the machine learning model.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 31, 2023
From: SUZUKI, KENJI
To: SONY GROUP CORPORATION
Reel/Frame 063807/0779 →
Priority Claims (1)
JP 2020-204550 · Dec 9, 2020 · national
Continuity (1)
Related Publication 20240005643A1 · Jan 4, 2024
References Cited (19)
US 10783401B1 · Jiang · 2020 [cited by examiner]
US 20190262084A1 · Roh · 2019 [cited by examiner]
US 20200250304A1 · Kruus et al. · 2020 [cited by applicant]
US 20210019641A1 · Park · 2021 [cited by examiner]
US 20210110439A1 · Jayne · 2021 [cited by examiner]
US 20210272016A1 · Veshchikov · 2021 [cited by examiner]
US 20210406712A1 · Bhide · 2021 [cited by examiner]
US 20220101146A1 · el Kaliouby · 2022 [cited by examiner]
JP 2018072893A · 2018 [cited by applicant]
JP 2019076699A · 2019 [cited by applicant]
JP 2019200769A · 2019 [cited by applicant]
JP 6635221B1 · 2020 [cited by applicant]
JP 2020534594A · 2020 [cited by applicant]
JP 6779491B1 · 2020 [cited by applicant]
International Search Report and Written Opinion of PCT Application No. PCT/JP2021/038146, issued on Dec. 14, 2021, 16 pages of ISRWO. [cited by applicant]
Toshiaki Inoue, “A scene recognition method using dashcams for reducing traffic accident risks”, Pioneer R&D, Oct. 31, 2020, 05 pages. [cited by applicant]
Hiroyuki Umeda, “Transfer learning and overfitting”, The Basics of AI Every Engineer Needs to Know: Easy Explanation of Machine Learning, Statistics, and Algorithms, Impress Corporation, Jan. 21, 2019, pp. 80-84. [cited by applicant]
Sun, et al, “Training Augmentation with Adversarial Examples for Robust Speech Recognition”, arxiv.org, Jun. 17, 2018, 05 pages. [cited by applicant]
Goodfellow, et al., “Explaining and Harnessing Adversarial Examples”, arxiv.org, Mar. 20, 2015, 12 pages. [cited by applicant]