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

Improved AdaBoost algorithm-based integrated deep belief network image identification method

A deep belief network and image recognition technology, applied in the field of deep learning and image recognition, can solve problems such as inability to process and recognize images, and achieve the effect of solving gradient dispersion and improving recognition rate.

Inactive Publication Date: 2017-11-07
JIANGNAN UNIV
View PDF3 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to human participation and guidance, traditional image recognition technology cannot continuously and efficiently process and recognize images, and can no longer meet the needs of practical applications. Therefore, the research and development of intelligent image recognition systems has great application value.

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
  • Improved AdaBoost algorithm-based integrated deep belief network image identification method
  • Improved AdaBoost algorithm-based integrated deep belief network image identification method
  • Improved AdaBoost algorithm-based integrated deep belief network image identification method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0023] (1) Preparation of training and test sets: Take 1000 images of each class from the MNIST dataset to form a training set, and 100 images of each class to form a test set; take 500 images of each class from the USPS dataset to form a training set, and 50 images of each class form a test set set; 300 images of each class are taken from the ETH-80 dataset to form a training set, and 100 images of each class form a test set.

[0024] (2) Construction of DBN model: In the implementation of the method, we consider the following four DBN network structure models: DBN(4)[400,256,100,64], DBN(6)[400,256,196,144,100,64] and AdaBoost+DBN(4){[400,256,100 , 64], [256, 144, 100, 64], [100, 81, 81, 60], [225, 196, 144, 100]}, where the data in the square brackets of each model is the number of hidden layer units of the model, and the number of data represents the hidden layer of the model. layers.

[0025] (3) Use the normalized image vector as the input of the model to train each net...

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 present invention provides an improved AdaBoost algorithm-based integrated deep belief network image identification method which solves the conflict between an image identification effect and a deep network scale to a certain extent, and improves the image identification efficiency. The method is mainly characterized by (1) separately training a plurality of parallel deep belief network (DBN) models of different scales and the corresponding classifiers; (2) utilizing an AdaBoost algorithm to integrate a plurality of DBN classifiers of different structures into a strong classifier; (3) carrying out the integral depth fine tuning on the parameters of the strong classifier via a BP algorithm, and obtaining a better local optimal solution by the iteration adjustment within the relatively shorter time, thereby being able to improve the classification effects of the models while reducing the errors.

Description

Technical field: [0001] The invention relates to the field of deep learning and image recognition, in particular to a method for recognizing images by using multiple parallel DBNs with different scales and independent of each other. Background technique: [0002] With the continuous development and progress of computer technology, people have more and more application requirements for digital images, such as military reconnaissance, safety in public places, industrial automation, life intelligence and environmental monitoring. With the rapid development of image capture devices (such as surveillance cameras, camera phones, sports cameras, etc.), massive image data can be acquired more easily, and the environment for data collection is easier to control; with the technological upgrade of storage devices, massive data The storage problem was solved, making it possible to significantly improve the performance of image recognition technology. However, traditional image recognit...

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): G06K9/62G06N3/08
CPCG06N3/084G06F18/2453G06F18/214
Inventor 陈秀宏田进
Owner JIANGNAN UNIV
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