Three-dimensional salient object detection technology of multi-attention guided neural network

A neural network and object detection technology, applied in the field of three-dimensional salient object detection, can solve the problems of misjudgment and affect the robustness of the algorithm, and achieve the effect of avoiding overfitting, improving misjudgment and reducing complexity.

Pending Publication Date: 2020-10-16
ZHEJIANG UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Our research found that such a principle error in the depth map will greatly affect the robustness of the algorithm and lead to misjudgment

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
  • Three-dimensional salient object detection technology of multi-attention guided neural network
  • Three-dimensional salient object detection technology of multi-attention guided neural network
  • Three-dimensional salient object detection technology of multi-attention guided neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

[0043] The three-dimensional salient object detection technology of a kind of multi-attention-oriented neural network that the present invention proposes, it comprises two processes of training phase and testing phase, and the concrete steps of described training phase process are:

[0044] Step 1_1: Select NJU2000 public data set and divide it into training set and test set; then use bilinear interpolation method to scale all images in the database to 224×224 size; Eighty of them are used as the training set, Indicates the kth color image, the color image is as follows figure 2 a. image 3 a and Figure 4 as shown in a; Then it represents the corresponding disparity map, and the disparity map is as follows figure 2 b. image 3 b and Figure 4 as shown in b; Indicates the corresponding label map; where k is a positive integ...

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 three-dimensional salient object detection technology of a multi-attention guided neural network. A double-flow neural network is constructed in a two-input mode, and an attention mask graph is generated by using two pieces of information, namely RGB and a disparity map; wherein the mask pattern comprises information of two modes, so that the two modes complement each other, and the principle problem of the disparity map can be avoided; the position information of the object can be quickly determined while the program running speed is increased; secondly, decoding isperformed by using down-sampling coding and up-sampling, and pixels of the image at detail positions are recovered step by step, so that the final accuracy is improved, and the result is closer to thecondition of a real human observation object. The negative effect brought by the principle problem that the pixel is still the highest when the object closest to the disparity map is a non-significant object is solved.

Description

technical field [0001] The invention relates to a stereo image processing technology based on binocular vision, in particular to a multi-attention oriented neural network stereo salient object detection technology. Background technique [0002] Human beings observe objects in order. The objects that people observe first are called salient objects, and the objects that people observe later are called non-salient objects. The algorithm technology that simulates the observation habit of human beings is the salient object detection technology. In recent years, this technology has a large number of applications in object classification, area suggestion, tracking, etc., and the technology has also continued to develop. There have been deep learning and three-dimensional salient object detection technologies. Although stereo saliency object detection techniques utilizing depth maps and disparity maps have achieved good performance, they still face many problems. [0003] Since th...

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): G06K9/62G06N3/04
CPCG06N3/045G06F18/214
Inventor 周武杰陈昱臻雷景生强芳芳王海江何成
Owner ZHEJIANG UNIVERSITY OF SCIENCE AND TECHNOLOGY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap