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

Graph attention target tracking method based on transfer learning-angular point prediction

A technology of target tracking and transfer learning, which is applied in the field of target tracking and graph attention target tracking based on transfer learning-corner prediction, which can solve the problems of inconvenient decoupling and inaccurate tracking results, and achieve fast speed, reduce parameters and calculations The effect of high volume and initial performance

Pending Publication Date: 2022-02-15
CHONGQING UNIV OF POSTS & TELECOMM
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

These two branches share the same response map output by GAM, forcing the same feature map to be classified and regressed, which is inconvenient for decoupling and leads to inaccurate tracking results

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
  • Graph attention target tracking method based on transfer learning-angular point prediction
  • Graph attention target tracking method based on transfer learning-angular point prediction
  • Graph attention target tracking method based on transfer learning-angular point prediction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0024] The technical solutions in the embodiments of the present invention will be described clearly and in detail below with reference to the drawings in the embodiments of the present invention. The described embodiments are only some of the embodiments of the invention.

[0025] The technical scheme that the present invention solves the problems of the technologies described above is:

[0026] The present invention utilizes the pre-trained GoogLeNet network to extract features from images, uses GAM to effectively embed information, and uses corner point prediction when predicting the position of a target frame, thereby avoiding uncertainty in coordinate prediction.

[0027] Combining with Figures 1 to 2, the present invention provides a method for tracking objects based on transfer learning-corner point prediction, which includes the following steps:

[0028] Step 1: Pre-train the GoogLeNet network, use the GOT-10k dataset as the training dataset, and train GoogLeNet to ha...

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 requests to protect a graph attention target tracking method based on transfer learning-angular point prediction. The method comprises the following steps: (1) pre-training a GoogLeNet network; (2) preprocessing a training set data set; (3) freezing characteristic layer parameters; (4) allowing characteristics to pass through a graph attention module (GAM); (5) remodeling a new feature sequence into feature mapping which serves as input of a full convolutional network (FCN), and outputting two probability graphs for the upper left corner and the lower right corner of an object bounding box respectively; (6) calculating an expected value of angular point probability distribution to obtain a prediction frame coordinate; (7) utilizing stochastic gradient to solve a loss function in advance, upgrading and optimizing parameters of a full connection layer, and finely adjusting weight parameters and bias parameters; and (8) calculating the position of a prediction frame by using angular point prediction, and framing a target. According to the method, better model parameters are obtained more quickly by utilizing transfer learning; and meanwhile, decoupling of classification and regression is realized by utilizing angular point prediction, a training process is simplified, and the speed of the whole network is increased.

Description

technical field [0001] The invention belongs to the technical field of image processing, and mainly relates to target tracking, in particular to a graph attention target tracking method based on transfer learning-corner point prediction. Background technique [0002] Visual object tracking refers to the detection, extraction, identification and tracking of moving objects in the image sequence, and obtains the moving parameters of the moving object, such as position, speed, acceleration and trajectory, so as to carry out the next step of processing and analysis, and realize the Behavioral understanding of moving objects for higher level detection tasks. In the tracking process, we hope to solve the problem of tracking failure caused by occlusion, uneven illumination, and deformation. [0003] In recent years, transfer learning has been widely used. Compared with traditional methods, transfer learning can take better hyperparameters of the original model, modify unsuitable p...

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): G06T7/20G06N3/04G06N3/08
CPCG06T7/20G06N3/08G06T2207/20081G06T2207/20132G06N3/047
Inventor 孙开伟冉雪宣立德李彦刘虎
Owner CHONGQING UNIV OF POSTS & TELECOMM
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