Image acquisition method and device, target identification method and device, and model training method and device

An acquisition method and model training technology, applied in the field of image processing, can solve problems affecting the reliability of image processing models, over-fitting of image processing models, and high training costs, so as to reduce training costs, alleviate over-fitting, and improve reliability effect

Active Publication Date: 2021-12-10
ZHEJIANG DAHUA TECH CO LTD
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] In the training tasks of image processing models, such as image segmentation, object recognition, text recognition, object classification, and object positioning, the samples that image processing models can include and process are limited, resulting in overfitting of image processing models and affecting image processing. The reliability of the model, and a large number of samples are included in the training process of the image processing model at the same time, and the training cost is relatively high

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 acquisition method and device, target identification method and device, and model training method and device
  • Image acquisition method and device, target identification method and device, and model training method and device
  • Image acquisition method and device, target identification method and device, and model training method and device

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach

[0040] Process the first image to obtain a positive sample set, that is, extract the target objects contained in multiple first images, and obtain positive samples by means of image matting, cropping, etc., and combine the positive samples to form a positive sample set.

Embodiment approach

[0042] The backgrounds other than the target object in the plurality of first images may be extracted through image matting, cropping, etc. to obtain negative samples, and the negative samples obtained from the first images are combined to form a first negative sample set. Moreover, since the negative samples in the first negative sample set are obtained in the first image, compared with the negative samples directly obtained in the second image, it can reflect the background characteristics when the target object exists, and has authenticity, and further Reduce model overfitting.

[0043] The backgrounds in the plurality of second images may be extracted through image matting, cropping, etc., to obtain negative samples, and the negative samples obtained from the second images are combined to form a second negative sample set. That is to say, in the sampling method of this embodiment, the source of positive samples is the first image, and the source of negative samples can be ...

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 discloses an image acquisition method and device, a target identification method and device, and a model training method and device. The image data acquisition method comprises the following steps: processing a first image to obtain a positive sample set; processing the first image and/or the second image to obtain a negative sample set, wherein the first image contains a target object, and the second image does not contain the target object; and performing sample combination based on the positive sample set and the negative sample set to obtain a plurality of different image subsets, wherein positive samples and/or negative samples in every two image subsets in the plurality of image subsets are different; and the plurality of image subsets are used for training the same image processing model. Through the above mode, the method can alleviate the overfitting of the image processing model, improves the reliability of the image processing model, and facilitates the reduction of the training cost.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to an image data acquisition method, a target recognition model training method, a target recognition method, electronic equipment and a computer-readable storage medium. Background technique [0002] In the training tasks of image processing models, such as image segmentation, object recognition, text recognition, object classification, and object positioning, the samples that image processing models can include and process are limited, resulting in overfitting of image processing models and affecting image processing. The reliability of the model, and a large number of samples are included in the training process of the image processing model at the same time, and the training cost is relatively high. Contents of the invention [0003] In view of this, the technical problem mainly solved by the present invention is to provide an image data acquisition method, an...

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): G06K9/62
CPCG06F18/22G06F18/214
Inventor 王超运殷俊潘华东孙鹤
Owner ZHEJIANG DAHUA 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