Information processing apparatus, information processing method, and non-transitory computer readable medium

Patent metadata

Field	Value
Publication number	US-11922677-B2
Application number	US-201917442757-A
Country	US
Kind code	B2
Filing date	Mar 27, 2019
Priority date	Mar 27, 2019
Publication date	Mar 5, 2024
Grant date	Mar 5, 2024

Abstract

An information processing apparatus ( 10 ) according to the present disclosure includes: an object recognition unit ( 11 ) that outputs, by using a first modal signal and a first modal recognition model corresponding to the first modal signal, an inference result regarding the first modal signal; a training data processing unit ( 12 ) that generates first modal training data regarding the first modal signal by using the inference result, and updates second modal training data regarding a second modal signal that is different from the first modal signal by using the first modal training data; and a recognition model update unit ( 13 ) that updates a second modal recognition model corresponding to the second modal signal by using the second modal training data.

First claim

What is claimed is:

1. An information processing apparatus comprising:
at least one memory storing instructions, and
at least one processor configured to execute the instructions to;
output, by using a first modal signal and a first modal recognition model corresponding to the first modal signal, an inference result regarding the first modal signal;
generate first modal training data regarding the first modal signal by using the inference result, and updating second modal training data regarding a second modal signal by using the first modal training data, the second modal signal being different from the first modal signal; and
update a second modal recognition model corresponding to the second modal signal by using the second modal training data.

2. The information processing apparatus according to claim 1 , wherein the second modal signal is image data acquired at substantially the same time that the first modal signal is acquired, the image data indicating substantially the same space as that of the first modal signal.

3. The information processing apparatus according to claim 2 , wherein the at least one processor is further configured to execute the instructions to generate area information indicating an area of a predetermined object included in the image data as the first modal training data, and use the area information as the second modal training data.

4. The information processing apparatus according to claim 3 , wherein the at least one processor is further configured to execute the instructions to
output a degree of reliability of the area information together with the area information, and
use the area information as the second modal training data when the degree of reliability exceeds a predetermined threshold.

5. The information processing apparatus according to claim 3 , wherein the at least one processor is further configured to execute the instructions to propagate coordinates of four vertices of a rectangle indicated by rectangle information to the second modal training data, and the rectangle information is the area information.

6. The information processing apparatus according to claim 3 , wherein the at least one processor is further configured to execute the instructions to propagate a mask image to the second modal training data, and the mask image is the area information.

7. The information processing apparatus according to claim 1 , wherein the at least one processor is further configured to execute the instructions to select either the first modal signal or the second modal signal and input a modal recognition model corresponding to the selected modal signal.

8. An information processing method comprising:
outputting, by using a first modal signal and a first modal recognition model corresponding to the first modal signal, an inference result regarding the first modal signal;
generating first modal training data regarding the first modal signal by using the inference result;
updating second modal training data regarding a second modal signal by using the first modal training data, the second modal signal being different from the first modal signal; and
updating a second modal recognition model corresponding to the second modal signal by using the second modal training data.

9. A non-transitory computer readable medium storing a program for causing a computer to:
output, by using a first modal signal and a first modal recognition model corresponding to the first modal signal, an inference result regarding the first modal signal;
generate first modal training data regarding the first modal signal by using the inference result;
update second modal training data regarding a second modal signal by using the first modal training data, the second modal signal being different from the first modal signal; and
update a second modal recognition model corresponding to the second modal signal by using the second modal training data.

Assignees

• Nec Corp

Inventors

• Miyano Hiroyoshi
• Toda Masato
• SAWADA Azusa
• Shibata Takashi

Classifications

• G06N3/0895

  Weakly supervised learning, e.g. semi-supervised or self-supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06V10/774 Primary

  Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06V10/87

  using selection of the recognition techniques, e.g. of a classifier in a multiple classifier system · CPC title