Check patentability & draft patents in minutes with Patsnap Eureka AI!

Data enhancement method, system and terminal

A technology for enhancing systems and data, applied in the field of computer vision, can solve the problems of complex data enhancement work and low work efficiency

Pending Publication Date: 2020-11-13
SHANGHAI INST OF MICROSYSTEM & INFORMATION TECH CHINESE ACAD OF SCI
View PDF11 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a data enhancement method, system and terminal, which are used to solve the problem that the data enhancement method in the prior art cannot meet the high performance of a large amount of input data for the model. training needs, and the complexity of data enhancement work leads to the problem of low work efficiency

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
  • Data enhancement method, system and terminal
  • Data enhancement method, system and terminal
  • Data enhancement method, system and terminal

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0056] Embodiment 1: A data augmentation method applied to a target tracking task.

[0057] The methods include:

[0058] Obtain the first frame of the video sequence to be tracked and the specific location and size information of the target to be tracked next;

[0059] Filtering one or more local pixels in the edge of the picture, and obtaining edge regions corresponding to each local pixel;

[0060] Set the size of the local disturbance area to 2ⅹ2, 3ⅹ3, and 4ⅹ4 in turn, the area mode is non-overlapping, and the disturbance mode is selected as random exchange of pixel positions in the local area, that is, all pixel values ​​​​are retained; data printing is performed on the pixels in the edge area. Randomly, obtain the enhanced samples corresponding to the image to be enhanced, input the tracking model to initialize the tracker and track the target in the next video sequence.

[0061] Among them, the tracker adopts ECO and Super_DiMP, and is tested on the GOT-10 and LaSOT d...

Embodiment 2

[0064] Embodiment 2: A data enhancement method applied to target detection and target segmentation tasks.

[0065] The methods include:

[0066] Obtain the first frame picture of the video sequence to be detected and segmented;

[0067] Filtering one or more local pixels in the edge of the picture, and obtaining edge regions corresponding to each local pixel;

[0068] Set the size of the local disturbance area to 2ⅹ2, the area mode is non-overlapping, and the disturbance mode is selected as random exchange of pixel positions in the local area; the data of the pixels in the edge area is disturbed, and the enhanced samples corresponding to the image to be enhanced are obtained ; Use the input target detection and target segmentation model to detect and track the target.

[0069] The target segmentation and detection model uses Mask R-CNN (Backbone: ResNet+FPN) and PointRend, and is tested on the COCO dataset to obtain a comparison list of target detection and segmentation indi...

specific Embodiment

[0074] Such as Figure 4 A schematic structural diagram of a system showing a data enhancement method in an embodiment of the present invention.

[0075] The system includes:

[0076] Input module 41, for inputting the image to be enhanced;

[0077] A local enhancement area module 42, connected to the input module 41, for screening one or more local pixels in the image to be enhanced, so as to obtain local enhancement areas corresponding to each local pixel according to preset enhancement area selection conditions ;

[0078] The data scrambling module 43 is connected to the local enhancement area module 42, and is used to perform data scrambling on one or more pixels in each local enhancement area based on the data scrambling standard, and obtain the image corresponding to the image to be enhanced. Enhanced samples.

[0079] Optionally, the image to be enhanced input by the input module 41 may be the original input image or an image that has undergone data enhancement, suc...

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 a data enhancement method and system and a terminal. The method comprises the steps: inputting a to-be-enhanced image; screening one or more local pixels in the to-be-enhancedimage so as to obtain a local enhancement region corresponding to each local pixel according to a preset enhancement region selection condition; based on a data disruption standard, respectively carrying out data disruption on one or more pixels in each local enhancement region to obtain an enhancement sample corresponding to the to-be-enhanced image. The problems that in the prior art, a data enhancement mode cannot meet the requirement for high-performance training of a large amount of input data of a model gradually nowadays, data enhancement work is complex, and consequently work efficiency is not high are solved. According to the method, the enhanced sample is obtained by disturbing the positions of the pixels in the region meeting a certain condition in the training sample within a certain range, so that the performance and generalization ability of the model are improved, and the working efficiency is improved. Therefore, the method effectively overcomes various defects in the prior art and has high industrial utilization value.

Description

technical field [0001] The present invention relates to the field of computer vision, in particular to a data enhancement method, system and terminal. Background technique [0002] Machine learning algorithms rely on data to train models. In recent years, with the continuous development of deep learning technology, the model has become more and more complex, the number of parameters has also increased, and the requirement for training data is increasing. Manual labeling is time-consuming and labor-intensive. This is when more training data can be generated through data enhancement, which reduces the possibility of model overfitting to a certain extent and improves the accuracy and generalization performance of the model. Until now, there have been various methods of data augmentation. Specifically, in the field of computer vision, image enhancement methods such as mirroring, random rotation, filtering, etc., but the above data enhancement methods cannot meet the needs of ...

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): G06T5/00G06T7/12G06T7/246G06N3/04G06N3/08
CPCG06T7/12G06T7/246G06N3/08G06T2207/10016G06N3/045G06T5/94
Inventor 谷宇章邱守猛袁泽强张晓林
Owner SHANGHAI INST OF MICROSYSTEM & INFORMATION TECH CHINESE ACAD OF SCI
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