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

Fast Face Retrieval Method Based on Triplet Deep Binary Network

A technology of deep network and binary network, applied in the field of image processing, can solve problems such as poor precision, and achieve the effect of solving low precision, strong representation ability, and accelerating training speed

Active Publication Date: 2021-07-02
TONGJI UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This method solves the problems of slow triplet training and poor accuracy of the two-stage method at the same time, achieving double improvement in accuracy and speed

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
  • Fast Face Retrieval Method Based on Triplet Deep Binary Network
  • Fast Face Retrieval Method Based on Triplet Deep Binary Network
  • Fast Face Retrieval Method Based on Triplet Deep Binary Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] The method scheme of the present invention is as follows as a whole: a series of face picture groups with training pictures reprocessed into triplets are given, firstly utilize the triplet hash coding network after coding vector grouping, and use the block diagram cutting method (existing Technical work, published in the paper "Fast Supervised Hashing with Decision Trees for High-Dimensional Data" (Lin G, Shen C, Shi Q, et al. Fast Supervised Hashing with Decision Trees for High-Dimensional Data[C] / / Computer Vision and PatternRecognition.IEEE,2014:1971-1978.) training to obtain the encoding vector of a certain group of bits. Then input the training picture into the deep network, extract the binary encoding feature of the last layer and compare it with the encoding vector obtained above , Feedback the result to the hash coding network, and repeat the above process. After repeated coding and feature training, a binary code with better representation ability is obtained. Fi...

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 proposes a method for fast face retrieval based on a triplet deep binary network, which belongs to the technical field of image processing. First, triplet preprocessing and encoding grouping are performed on the image, and then triplet hash encoding training is performed using the block image cutting method, feature extraction is performed using a deep network, and a loop interleaving two-stage method is used to effectively feed back the discrimination information of the deep network to the The hash coding network enables the two stages to carry out cyclic learning and mutual correction, and finally extract the discriminative features of the deep network as the hash code of the picture for feature comparison and face retrieval. In this way, block coding training can be performed to speed up training, and at the same time, deep network information can be effectively used to extract highly discriminative hash codes and improve retrieval accuracy.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a fast face retrieval method based on a triplet deep binary network. Background technique [0002] Face retrieval refers to finding samples of the same person as the face to be recognized from the huge face database, that is, to confirm whether the face in the database and the face to be recognized are the same person one by one. This problem has important practical value in the fields of access control facial recognition, video surveillance, and facial recognition payment. [0003] With the rapid development of big data, face retrieval is greatly limited in speed. In order to ensure the speed, the features are usually encoded with binary values. Binary hash coding maps the original features to compressed binary codes, thus speeding up feature matching. The two-stage hash coding method separates coding training and matrix learning, effectively improving the training efficiency ...

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 Patents(China)
IPC IPC(8): G06F16/583G06N3/04G06N3/08
CPCG06F16/583G06N3/08G06N3/045
Inventor 尤鸣宇沈春华张欣彧
Owner TONGJI UNIV
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com