Unlock instant, AI-driven research and patent intelligence for your innovation.

Multi-label image recognition method and device

An image recognition device and multi-label technology, applied in the field of computer vision, can solve the problems of model memory and speed impact, model complexity, poor robustness, etc., to achieve the effect of improving accuracy, good robustness, and improved accuracy

Active Publication Date: 2020-11-13
SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF12 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This method will inhibit the categories with low co-occurrence frequency, so it is difficult to further improve the recognition accuracy of multi-label classification through this static image method
[0006] 2), the existing static graph construction method must perform probability statistics on the data set in advance, which makes the model more complex and less robust
[0007] 3) The existing multi-label classification method is too complicated to separate the feature map of the image, resulting in a relatively large impact on the memory and speed of the model

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Multi-label image recognition method and device
  • Multi-label image recognition method and device
  • Multi-label image recognition method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangements of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0028] The following description of at least one exemplary embodiment is merely illustrative in nature and in no way taken as limiting the invention, its application or uses.

[0029] Techniques, methods and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such techniques, methods and devices should be considered part of the description.

[0030] In all examples shown and discussed herein, any specific values ​​should be construed as exemplary only, and not as limitations. Therefore, other instances of the exemplary embodiment may have dif...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a multi-label image recognition method and device. The device comprises: a semantic attention module used for separating a feature map output by a backbone network into a plurality of categories of features; and a dynamic graph convolutional network module which is used for modeling a relationship among the plurality of category features by using a dynamic graph convolutional network, wherein the dynamic graph convolutional network comprises a static graph and a dynamic graph, the static graph is used for acquiring global correlation of the image, and the dynamic graphis used for acquiring local correlation of the image. The method and device can improve the accuracy of image recognition, is higher in independence and robustness, and can be used for the image recognition of various types of scenes.

Description

technical field [0001] The present invention relates to the technical field of computer vision, and more specifically, to a multi-label image recognition method and device. Background technique [0002] In recent years, Graph Neural Networks (GNN) have been widely used in computer vision or NLP (Natural Language Processing). The graph neural network obtains the correlation between different nodes by modeling the relationship between all node features, improves the expressive ability of node features, and then improves the accuracy of the target task. The most basic unit of a graph neural network consists of a relational modeling layer and a state update layer. Usually, a graph neural network consists of n (n>=1) basic units. For the relationship modeling layer, the general practice is to use a graph (graph) to model the relationship between nodes. Depending on the state update layer, the naming of the graph neural network is also different. For example, if the state up...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06K9/46G06N3/04
CPCG06V10/40G06N3/045G06F18/24
Inventor 乔宇彭小江叶锦
Owner SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI