Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Saliency object detection method based on deep convolutional network

A technology of deep convolution and target detection, applied in the field of target detection

Active Publication Date: 2017-12-01
NAT UNIV OF DEFENSE TECH
View PDF4 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is to improve the robustness of the saliency extraction method and reduce the use of artificially designed specific feature description methods

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
  • Saliency object detection method based on deep convolutional network
  • Saliency object detection method based on deep convolutional network
  • Saliency object detection method based on deep convolutional network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] The implementation process of this method is divided into two stages: training and testing of salient object detection. The specific implementation of this method will be described below.

[0028] The present invention adopts the ECSSD and MSRA10K databases that are widely used in the field of saliency detection as research objects (both databases disclose their real saliency detection values ​​and original data), among which MSRA10K is the largest saliency library released so far , containing 10,000 image sources; ECSSD is a semantically rich but complex database, containing 1,000 image sources. In the training phase, 95% of the data in the two databases were selected as the source of the basic training set, and the remaining 5% of the data was used in the test set.

[0029] (1) Saliency detection training process of deep network

[0030] Step 1. According to the screened 95% original image data, label saliency map, and training data set construction method, randomly...

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 belongs to the field of object detection and discloses a saliency detection method based on a deep convolutional neural network. The method comprises the steps of (1) network training data construction, that is, to construct a training image block data sample set according to a given image data set and a marked saliency map thereof; (2) database pre-processing, that is, to preprocess pixels of each image block data according to the constructed training database; (3) network structure design, that is, to extract salient objects in image blocks through the design of a deep network structure (referring in particular to I[28x28x3]-C[24x24x20]-P[12x12x20]-C[8x8x50]-P[4x4x50]-FC[500)]-O[1]); and (4) network structure training, that is, to update a deep network model by calculating an error function using the difference between the output of the deep convolutional network and label data. The method provided by the invention has strong robustness and does not require manual design of a specific feature description mode.

Description

Technical field: [0001] The present invention mainly relates to the field of target detection, in particular to a salient target detection method based on a deep convolutional network. Background technique: [0002] Inspired by the ability of human vision to perceive the external environment, saliency detection algorithms have become a research hotspot in the field of vision in recent years. At present, the saliency detection technology is not mature enough. In addition to the performance of the saliency algorithm itself is not high enough, the application method of saliency information is not perfect enough, and a more satisfactory implementation method needs to be found. In the past 10 years, deep learning has achieved great success in speech recognition, natural language processing, computer vision, image and video analysis, multimedia and many other fields, and has become one of the important branches of artificial intelligence. This patent intends to use deep learning ...

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
CPCG06N3/04G06V10/462G06F18/211G06F18/214
Inventor 牛轶峰马兆伟王菖赵哲
Owner NAT UNIV OF DEFENSE TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com