Online behavior recognition model training and detection method and system

A technology for identifying models and detection methods, applied in the field of computer vision, can solve problems such as difficult to optimize architecture, unsatisfactory performance, non-parallelism and gradient disappearance, and achieve the effect of reducing network parameters and calculation load

Pending Publication Date: 2022-01-28
JIANGSU ELECTRIC POWER CO +1
View PDF3 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, RNN-like architectures suffer from non-parallelism and vanishing gradients
Therefore, it is difficult to optimize the architecture, which may lead to suboptimal performance
This is a challenging problem for current methods

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
  • Online behavior recognition model training and detection method and system
  • Online behavior recognition model training and detection method and system
  • Online behavior recognition model training and detection method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0073] The application will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solutions of the present invention more clearly, but not to limit the protection scope of the present application.

[0074] Such as figure 1 As shown, Embodiment 1 of the present invention provides an online behavior recognition model training and detection method, comprising the following steps:

[0075] Step 1: Input the input video stream to the spatial Transformer feature extraction network of the online behavior recognition model to output the spatial features representing the visual features of each frame. Specifically include:

[0076] The input video stream V is represented by the following formula,

[0077]

[0078] In the formula:

[0079] f t Indicates the video frame at time t,

[0080] T means t 0 T times before time,

[0081] That is, the input video stream V is represented by f -T ,...,...

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 online behavior recognition model training and detection method and system. The method comprises the following steps: 1 inputting an input video stream into a spatial Transformer feature extraction network of an online behavior recognition model to output spatial features representing visual features of each frame; 2 constructing a token feature sequence based on the spatial features; 3 inputting the token feature sequence obtained in the step 2 into a Transformer model, identifying a behavior of a current frame block f0 by using an encoder of the Transformer model, and predicting an upcoming future behavior by using a decoder; 4 calculating the final training Loss of the whole behavior recognition model, implementing an offline training process, and obtaining an online behavior recognition model after the training is finished; and 5 after the above steps are finished, when an online video is input, the online behavior recognition model can output the behavior category of the current frame. According to the method, an online behavior recognition detection algorithm based on Transformer is innovatively adopted, and the online behavior real-time detection task is realized in the early stage of ensuring the accuracy.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to an online behavior recognition model training and detection method and system. Background technique [0002] Online action recognition is a computer vision task that correctly recognizes ongoing human actions from video streams. Online behavior recognition technology is different from traditional computer vision tasks. This technology emphasizes that the algorithm to realize this technology must have certain real-time performance while correctly recognizing ongoing human actions in video streams. This determines that online behavior recognition has two major technical difficulties: First, unlike image-based vision tasks, online behavior recognition tasks need to detect actions with insufficient observations when video frames arrive, which not only requires learning each video frame More importantly, it is necessary to fully exploit the temporal features b...

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): G06V40/20G06V20/40G06V10/77G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/2136
Inventor 崔隽峰张文彬张军民王东林席晓强李海冰刘晨张国梁吴鹏杜泽旭
Owner JIANGSU ELECTRIC POWER CO
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