Vulgar action recognition method

An action recognition, vulgar technology, applied in the field of action recognition, can solve the problems of low accuracy and timeliness of the recognition model, achieve the effect of increasing memory consumption and reducing the impact of video blur

Inactive Publication Date: 2022-03-08
北京智视数策科技发展有限公司
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] To this end, the embodiment of the present invention provides a vulgar action recognition method to solve the problem of low accuracy and timeliness of the recognition model in the prior art

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
  • Vulgar action recognition method
  • Vulgar action recognition method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0040] Embodiment: a kind of vulgar action recognition method, such as figure 1 As shown, the model needs to be trained before the video is recognized. In this embodiment, the training model must first determine the action classification standard, manually review the vulgar video, classify the video according to the content, and then cut the video into specified lengths. Video clips, and then use the classified processed video clips to train to obtain the network model. The specific strategy configuration of the training model: use the Momentum optimization algorithm for training, and introduce exponentially weighted moving average in the ordinary gradient descent method, that is, define a momentum (the index of the gradient Weighted moving average), and then use this value instead of the original gradient direction to update, where momentum=0.9; using L2_Decay, in the initial stage of training, because the weight is in a random initialization state, the loss function drops fas...

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 embodiment of the invention discloses a vulgar action recognition method which comprises the following steps: S1, judging whether a video needing to be detected is a short video or not, if the judgment result is yes, performing the next step, and if the judgment result is no, automatically and equally cutting the video needing to be detected into a plurality of segments, and performing the next step; s2, extracting a preset frame number of images from each segment of the plurality of segments to obtain an image frame sequence; s3, substituting the image frame sequence into a training model for processing, and then outputting apparent time sequence information; s4, processing the apparent time sequence information and then outputting a feature vector; and S5, comparing the feature vector with a preset vector threshold, and judging whether the video is a vulgar video or not, thereby realizing timely, efficient, time-saving and labor-saving monitoring of the video, live broadcast and other contents.

Description

technical field [0001] The embodiment of the present invention relates to the technical field of action recognition based on deep learning, and specifically relates to a vulgar action recognition method. Background technique [0002] With the rapid development of the mobile Internet, the explosive growth of online audio-visual data, and various forms of audio-visual content are produced and disseminated every day, such as long videos, short videos, live broadcasts, etc. Among them, the live broadcast content is more real-time and has a large amount of content, which brings great challenges to content supervision and investigation. Moreover, the common problems in the live broadcast supervision industry are that there are many platforms, large amount of data, high cost of manual review, and low efficiency. It is for these reasons that a large number of vulgar videos appear in our network environment. Timely discovery of such vulgar content, and its supervision, so that it se...

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): G06V20/40G06N3/04G06N3/08
CPCG06N3/08G06N3/045
Inventor 王金水
Owner 北京智视数策科技发展有限公司
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