Learning device, object detection device, learning method, and recording medium

Patent metadata

Field	Value
Publication number	US-12046000-B2
Application number	US-201917618550-A
Country	US
Kind code	B2
Filing date	Jun 25, 2019
Priority date	Jun 25, 2019
Publication date	Jul 23, 2024
Grant date	Jul 23, 2024

Abstract

A learning device makes an object detection device learn how to detect an object from an input image. A feature extraction unit performs feature extraction from input images including real images and pseudo images to generate feature maps, and the object detection unit detects objects included in the input images based on the feature maps. The domain identification unit identifies the domains forming the input images and generates domain identifiability information. Then, the feature extraction unit and the object detection unit learn common features that do not depend on the difference in domains, based on the domain identifiability information.

First claim

What is claimed is:

1. A learning device comprising:
a memory storing instructions; and
one or more processors configured to execute the instructions to:
perform feature extraction from input images including real images and pseudo image to generate feature maps;
detect objects included in the input images based on the feature maps; and
identify domains forming the input images; and
output domain identifiability information,
wherein the processor is configured to learn common features that do not depend on difference in domains, based on the domain identifiability information.

2. The learning device according to claim 1 ,
wherein the processor is configured to generate frame information indicating frames set on the feature map,
wherein the processor is configured to detect the object for each frame, and
wherein the processor is configured to identify the domain for each frame.

3. The learning device according to claim 2 , wherein the processor is configured to identify which one of a domain of real image data and a domain of pseudo image data an area defined by the frame of the input image corresponds to.

4. The learning device according to claim 3 , the processor is further configured to generate domain area information indicating an area of pseudo image data in the pseudo image based on attribute information of the pseudo image,
wherein the processor is configured to generate correct answer data of domain identification results for each frame based on the domain area information, and perform learning of domain identification using the correct answer data.

5. The learning apparatus according to claim 4 , wherein the processor is configured to generate the correct answer data of the domain identification result based on a positional relationship between the area of pseudo image data in the pseudo image and an area defined by the frame of the input image.

6. The learning device according to claim 1 , wherein the domain identifiability information is information indicating a difference between the domain of real image data and the domain of pseudo image data.

7. The learning device according to claim 1 , wherein the pseudo image is created by synthesizing the real image data and the pseudo image data.

8. An object detection device comprising a feature extraction unit and an object detection unit learned by the learning device according to claim 1 .

9. A learning method of an object detection device for extracting a feature from an input image and detecting an object included in the input image, comprising:
inputting input images including real images and pseudo images to the object detection device;
identifying domains forming the input images to generate domain identifiability information; and
making the object detection device learn common features that do not depend on difference in domains, based on the domain identifiability information.

10. A non-transitory computer-readable recording medium recording a program to execute learning processing of an object detection device including a computer, the object detection device extracting a feature from an input image and detecting an object included in the input image, the program causing the computer to:
input input images including real images and pseudo images to the object detection device;
identify domains forming the input images to generate domain identifiability information; and
make the object detection device learn common features that do not depend on difference in domains, based on the domain identifiability information.

Assignees

• Nec Corp

Inventors

• Ogawa Takuya
• Shibata Takashi
• Suzuki Tetsuaki
• Miyano Hiroyoshi

Classifications

• G06T2207/20081

  Training; Learning · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06V10/7715

  Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods · CPC title

• G06V10/25

  Determination of region of interest [ROI] or a volume of interest [VOI] · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title