Label adding method and device and terminal

A label and target image technology, applied in the field of image processing, can solve the problems of long label adding cycle, large labor cost, consumption, etc., and achieve the effect of improving label adding efficiency, saving time and saving human resources.

Inactive Publication Date: 2018-09-21
BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
View PDF3 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The embodiment of the present invention provides a tag adding method, device and terminal to solve the problems in the prior art that the tag adding cycle is long and consumes a lot of labor costs

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
  • Label adding method and device and terminal
  • Label adding method and device and terminal
  • Label adding method and device and terminal

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0027] refer to figure 1 , shows a flow chart of steps of a tag adding method according to Embodiment 1 of the present invention.

[0028] The tag adding method of the embodiment of the present invention may include the following steps:

[0029] Step 101: Search for a first preset number of seed images from an image library.

[0030] The seed image matches the first tag to be added. In the embodiment of the present invention, there is an image library and a pre-trained first image classification model in advance. The first image classification model includes a plurality of second labels and training images corresponding to each second label, forming a label system. When pre-adding a new label, namely the first label, to the first image classification model, it is first necessary to obtain the training image of the first label, and secondly, add the first label and the training image of the first label to the label system, and retrain The image classification model can compl...

Embodiment 2

[0043] refer to figure 2 , shows a flow chart of steps of a tag adding method according to Embodiment 2 of the present invention.

[0044] The tag adding method of the embodiment of the present invention may specifically include the following steps:

[0045] Step 201: Search for a first preset number of seed images from an image library.

[0046] The seed image matches the first tag to be added. In the embodiment of the present invention, there is an image library and a pre-trained first image classification model in advance. The first image classification model includes a plurality of second labels and training images corresponding to each second label, forming a label system. The first image classification model may be an open source image classification model. The first label to be added can be a single label, or two or more labels. The specific process for adding each label is the same, so in the embodiment of the present invention, a single first label is added to the...

Embodiment 3

[0070] refer to image 3 , shows a structural block diagram of a label adding device according to Embodiment 3 of the present invention.

[0071] The tag adding device of the embodiment of the present invention may include: a search module 301, configured to search for a first preset number of seed images from an image library; wherein, the seed images match the first tag to be added; the screening module 302 is configured to Configured to screen feature-similar images corresponding to each of the seed images from the image library according to the pre-trained first image classification model; wherein, each seed image corresponds to a second preset number of feature-similar images; generating module 303, configured to combine the feature similar images corresponding to each of the seed images to generate a training image of the first label; the training module 304 is configured to use the first label, the first The training image of the label, each second label included in th...

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

Embodiments of the invention provide a label adding method and device and a terminal. The method comprises the following steps of: searching a first preset quantity of seed images from an image library, wherein the seed images are matched with a to-be-added first label; screening feature similar images corresponding to the seed images from the image library according to a pre-trained first image classification model; summarizing the feature similar images corresponding to the seed images to generate a training image of the first label; and training a second image classification model accordingto the first label, the training image of the first label, second labels included in the first image classification model and training images of the second labels. Through the label adding method provided by the invention, users do not need to train image signs in batches, and only need to manually screen the first preset quantity of seed images, so that the operation is convenient, the consumedtime is short, the human resources can be saved and the label adding efficiency can be enhanced.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a tag adding method, device and terminal. Background technique [0002] Recently, deep learning has made breakthrough progress in natural language processing, text translation and other related content understanding fields. However, these developments are heavily dependent on the size of the training data, so data is the main bottleneck in applying these techniques to the actual production environment. [0003] Taking the current image classification task as an example, generally the amount of data required for each label is on the order of "thousands". The traditional method is to collect a sufficient amount of labeled data, that is, labeled images, and then add these labeled data to the existing classification system, but this approach has the following shortcomings: [0004] First, the "thousand" level of data corresponding to a label does not seem to be muc...

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): G06F17/30G06K9/62
CPCG06F18/214
Inventor 张志伟杨帆
Owner BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap