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

Attention-based blocking target detection method

A technology of target detection and attention, applied in neural learning methods, instruments, biological neural network models, etc., can solve problems such as slowing down

Inactive Publication Date: 2017-06-13
CHINA CHANGFENG SCI TECH IND GROUPCORP
View PDF3 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In target detection, methods such as selective search are mainly used to extract candidates for possible target areas, and then input them into the convolutional neural network. Since each picture needs to generate thousands of candidates, each candidate is then input into the convolutional neural network. This greatly slows down

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

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0012] Applying a convolutional neural network to process large-sized pictures requires a huge amount of calculation. The human eye processes pictures based on attention, so that there is key recognition, rather than processing the key points of the entire part of a picture seen. At the same time, the human learning process is a process of continuously strengthening correct understanding and gradually reducing mistakes. Enhanced learning is to imitate this principle.

[0013] First of all, for a picture, determine the focus of attention, randomly designate it for the first time, find a small area of ​​interest, and then create three proportionally sized picture blocks based on this, and then scale them to the same size, which is a glimpse of the human eye. Input the obtained three pictures into the recurrent neural network, and then the recursive network produces two outputs, one output enters the positioning network to generate positioning information Lt, which is used to dete...

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 an attention-based blocking target detection method. The method comprises the steps of firstly determining an attention focus for a picture, randomly specifying the picture at first, finding a small region of interest, then creating three picture blocks with proportional sizes by taking the small region of interest as a center, and then zooming the picture blocks to a same size; inputting three obtained pictures to a recurrent neural network, then generating two outputs by the recurrent network, and inputting one output to a locating network to generate locating information for re-determining a target focus of interest in the picture; and inputting the other output to a fully-connected network for determining whether a currently generated picture block is an object or not, and if yes, obtaining feedback of 1, otherwise, obtaining the feedback of 0 and serving as a reinforcement learning signal.

Description

technical field [0001] The invention belongs to the technical field of image information data processing, and relates to the application of image information data processing technology in the fields of deep learning, video analysis and target detection. Background technique [0002] The deep convolutional neural network has achieved the best results in the field of object detection and recognition. It is designed to process multi-dimensional array data and uses 4 key ideas to take advantage of the properties of natural signals: local connection, weight sharing, pooling and multiple use of the network layer. [0003] In target detection, methods such as selective search are mainly used to extract candidates for possible target areas, and then input them into the convolutional neural network. Since each picture needs to generate thousands of candidates, each candidate is then input into the convolutional neural network. This greatly slows down the speed. [0004] Google's de...

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
IPC IPC(8): G06K9/62G06N3/04G06N3/08
CPCG06N3/04G06N3/08G06F18/2111
Inventor 钟南成健张建伟张丹普张晓林王亚静
Owner CHINA CHANGFENG SCI TECH IND GROUPCORP
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