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

Unsupervised multi-source field adaptive method based on deep joint semantics

A multi-source field, unsupervised technology, applied in neural learning methods, instruments, biological neural network models, etc., can solve the problems of lack of category information and comprehensive consideration of context information, and achieve the effect of reducing cross-domain differences

Pending Publication Date: 2022-07-29
SHANDONG UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, these methods lack comprehensive consideration of category information and context information

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
  • Unsupervised multi-source field adaptive method based on deep joint semantics
  • Unsupervised multi-source field adaptive method based on deep joint semantics
  • Unsupervised multi-source field adaptive method based on deep joint semantics

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0034] This example provides an unsupervised multi-source domain adaptive network based on deep learning for image classification, which specifically includes the following steps:

[0035] Step 1: Obtain training sample images from different sources;

[0036] To verify the performance of our proposed multi-source domain adaptation method, we select the Office-31 dataset as our sample images from the domain adaptation open competition and select 2 datasets as the source domain and the other dataset as the target area. The Office-31 dataset is a domain-adaptive benchmark dataset with 4110 images and three subsets with different distributions, each containing 31 categories: Amazon(A) downloaded from amazon.com website, Webcam(W) 795 web cameras, DSLR (D) 498 digital SLR cameras. In addition, office-31 is an unbalanced dataset, each domain contains a different number of images, three transfer tasks are implemented here: A, W→D; W, D→A; D, A→W are used for Test the performance o...

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 relates to the technical field of image classification, in particular to an unsupervised multi-source field adaptive method based on a deep joint semantic maximum mean difference metric index, which is characterized by comprising the following steps: acquiring sample images from different sources, and preprocessing data; an unsupervised multi-source field adaptive network model is established, a joint semantic maximum average difference JSMMD is introduced to align cross-domain features learned by conditional distribution and joint distribution, and the unsupervised multi-source field adaptive network model comprises a common feature extractor, a task specific layer and a task classification layer; training an unsupervised multi-source field adaptive model; and obtaining a classification result of a target domain sample by using the trained unsupervised multi-source domain adaptive model.

Description

Technical field: [0001] The invention relates to the technical field of image classification, in particular to an unsupervised multi-source domain self-adaptation method based on a deep joint semantic maximum mean difference measurement index. Background technique: [0002] Deep learning has been widely used in computer vision fields such as image classification and object detection, and has achieved remarkable results. The availability of large-scale labeled datasets plays a crucial role in the great success of deep learning. However, the manual labeling process is very time consuming and laborious. How to use the existing labeled domain data to solve the corresponding tasks in the unknown related unlabeled domain is an important direction to solve large-scale data annotation. [0003] In recent years, Unsupervised Domain Adaptation (UDA) has emerged as an attractive solution to the data labeling problem by transferring information from an existing labeled source domain t...

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): G06V10/774G06K9/62G06V10/764G06V10/80G06V10/82G06F16/953G06N3/04G06N3/08
CPCG06F16/953G06N3/088G06N3/045G06F18/2155G06F18/2431G06F18/2415G06F18/259G06F18/253
Inventor 王帅程志明颜成钢
Owner SHANDONG 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