Image training sample mining method, device, terminal and computer-readable storage medium

A technology for training samples and pictures, which is applied in the field of information processing, can solve problems such as difficulty in obtaining pictures, limited performance of picture retrieval systems, and less public use of data user privacy and data privacy, so as to reduce labor costs, efficiently customize needs, and improve The effect of production efficiency

Active Publication Date: 2019-12-20
BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The accuracy of the training samples is high, but due to the limited performance of the image retrieval system, high-quality training samples cannot be guaranteed
User click behavior is an effective method to improve accuracy, but due to user privacy issues and data privacy, this data is rarely used publicly, and it is difficult to obtain it even within the same company

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
  • Image training sample mining method, device, terminal and computer-readable storage medium
  • Image training sample mining method, device, terminal and computer-readable storage medium
  • Image training sample mining method, device, terminal and computer-readable storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0068] In a specific embodiment, such as figure 1 As shown, a method for mining image training samples is provided, including:

[0069] Step S100: Obtain a plurality of candidate pictures and corresponding picture description texts according to the input picture query conditions.

[0070] The picture query condition may be a text query type entered through a text input box provided by the search engine, and the picture query condition may also be in the form of a picture. Preliminary screening of image training samples is performed according to the input image query conditions to obtain a set of candidate image samples. The candidate picture sample set includes multiple candidate pictures and picture description texts used to describe the candidate pictures. Wherein, the picture description text may be a picture text title or an artificially added semantic description, which is artificially labeled information of candidate pictures. The establishment of candidate image samp...

Embodiment 2

[0106] In a specific embodiment, such as Figure 7 As shown, a kind of image training sample mining device is provided, comprising:

[0107] A candidate picture acquisition module 10, configured to acquire a plurality of candidate pictures and corresponding picture description texts according to the input picture query condition;

[0108] The general text similarity model training module 20 is used to obtain the general text similarity model according to the picture description text training;

[0109] The vertical class model training module 30 is used for utilizing the general text similarity model and category feature parameter training to obtain the vertical class model, and the category feature parameter is corresponding to the training sample category obtained according to the picture description text classification;

[0110] A candidate picture classification module 40, configured to classify the candidate pictures using a vertical class model to obtain a plurality of c...

Embodiment 3

[0129] An embodiment of the present invention provides a picture training sample mining terminal, such as Figure 9 shown, including:

[0130] A memory 400 and a processor 500 , the memory 400 stores computer programs that can run on the processor 500 . When the processor 500 executes the computer program, the method for mining image training samples in the foregoing embodiments is implemented. The number of memory 400 and processor 500 may be one or more.

[0131] The communication interface 600 is used for the memory 400 and the processor 500 to communicate with the outside.

[0132] The memory 400 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory.

[0133] If the memory 400, the processor 500, and the communication interface 600 are implemented independently, the memory 400, the processor 500, and the communication interface 600 may be connected to each other through a bus to c...

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 proposes a picture training sample mining method, device and terminal, the method comprising: obtaining a plurality of candidate pictures and corresponding picture description texts according to input picture query conditions; training a general text similarity model according to the picture description text; Utilize the general text similarity model and category characteristic parameter training to obtain the vertical class model, the category characteristic parameter corresponds to the training sample category obtained according to the picture description text classification; use the vertical class model to classify the candidate pictures, and obtain a plurality of candidate pictures Classification set: Input the pictures in each candidate picture classification set into the text semantic similarity model and the picture content similarity model to obtain the picture training samples corresponding to each category. When the image query conditions are given, it can effectively and automatically mine image training samples, reduce labor costs, meet the customization needs of different customers, and improve the production efficiency of training samples.

Description

technical field [0001] The present invention relates to information processing technology, in particular to a method, device, terminal and computer-readable storage medium for mining picture training samples. Background technique [0002] The maturity of computer vision technology has brought breakthroughs in the fields of image classification, image retrieval, video analysis, video or image advertising, automatic driving, and intelligent medical care. In order to achieve higher image classification retrieval accuracy and image recognition rate, it is necessary to collect data for different application scenarios when training the visual model, such as flower recognition, red wine recognition, animal recognition, dog recognition, etc. At the same time, in order to increase the generalization ability of the visual model, a large number of image training samples are required. [0003] Currently, there are three options for mining image training samples: (1) Full manual labelin...

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): G06K9/62
CPCG06F18/22G06F18/24G06F18/214
Inventor 孟骧龙严灿祥
Owner BAIDU ONLINE NETWORK TECH (BEIJIBG) 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