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

Auxiliary data labeling method capable of achieving online learning

A technology of auxiliary data and data, which is applied in the field of computer vision and deep learning, can solve problems that consume a lot of manpower and time, and achieve the effect of improving accuracy, improving performance, and reducing time and labor costs.

Active Publication Date: 2019-08-23
TIANJIN UNIV +1
View PDF3 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to solve the technical problem that the existing data labeling process needs to be repeated and consume a lot of manpower and time. The present invention provides an auxiliary data labeling method that can be learned online

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
  • Auxiliary data labeling method capable of achieving online learning
  • Auxiliary data labeling method capable of achieving online learning
  • Auxiliary data labeling method capable of achieving online learning

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach

[0016] Specific implementation: This implementation is a method for annotating auxiliary data that can be learned online, and the specific steps are as follows:

[0017] 1) Using the initially marked data, train the model once to get M 1 : The deep target detection network faster rcnn is used during training, and the stochastic gradient descent method is used when training faster rcnn; the initial learning rate is set to 0.001 when the model is trained for the first time, and the initial learning rate is set to 0.0001 for subsequent training; each training When using 20% ​​of the data as the test set data;

[0018] 2) Judging whether there is new data to be marked, if there is new data to be marked, repeat the iterative calculation of step 3) to step 5), until there is no new data to be marked, the method ends;

[0019] 3) For the lth batch of data x that needs to be labeled l , using the model M obtained from the previous training l-1 Make predictions on the data: When p...

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 auxiliary data labeling method capable of achieving online learning. The method comprises the steps of first-time model training, model marking data using, marking data manual correction and model re-training optimization. According to the auxiliary data labeling method disclosed by the invention, model training can be completed under the condition of few initial test data, and the model is used for auxiliary data labeling; most importantly, the online learning is realized by training the model again by using the data marked by the auxiliary label each time.The performance of the model is further improved, and the accuracy of target detection of the model is improved, so that the accuracy of auxiliary data annotation is improved, and the consumption of time and labor cost in a repeated manual data annotation process is greatly reduced.

Description

technical field [0001] The invention relates to the fields of computer vision and deep learning, in particular to an online learning auxiliary data labeling method. Background technique [0002] Target detection is a very important branch in the field of computer vision. Through the target detection model machine, we can obtain the attributes of the object of interest in the picture, such as the area, category, and confidence. In order to make the target detection achieve a high accuracy rate, the target must be The detection model is trained. At present, models in the field of deep learning require a large amount of data for training. Since the data labels required for target detection have many attributes, the training of the current model is to manually label a large amount of data, use the labeled data to train the model, and label the data. This repetitive work requires a lot of manpower and time, and before the data training is completed, the model cannot be used for ...

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
IPC IPC(8): G06K9/62
CPCG06F18/40G06F18/2413Y02T10/40
Inventor 胡清华吴浩然温泉宝鹤鹏赵帅陈超李敏
Owner TIANJIN 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