Behavior recognition system from bottom to top and from top to bottom

A top-down, bottom-up technology, applied in character and pattern recognition, instruments, biological neural network models, etc., can solve poor performance, ignore long-range dependencies, and do not consider the influence of background information and surrounding information and other issues to achieve the effect of performance improvement and sensitivity improvement

Inactive Publication Date: 2019-06-07
GUANGZHOU INTELLIGENT CITY DEV INST +1
View PDF2 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The existing technical solution SHN uses intermediate supervision information to consider the problem of single-person gesture recognition. The design of the module mainly limits and solves the problem of gesture recognition, and does not consider the influence of background information and surrounding information in behavior recognition. Poor performance in human scenes
[

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
  • Behavior recognition system from bottom to top and from top to bottom
  • Behavior recognition system from bottom to top and from top to bottom
  • Behavior recognition system from bottom to top and from top to bottom

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0022] In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

[0023] refer to Figure 1 to Figure 3 ,Such as Figure 1 to Figure 3 As shown, a bottom-up-top-down behavior recognition system includes SBTA module and STBTA module; the SBTA module and STBTA module pass bottom-up top-down mechanism and attention mechanism to local features and Global information is encoded;

[0024] The bottom-up and top-down mechanism is to downsample the feature map layer by layer and then upsample layer by layer, preserve multi-scale learning through residual connection, and have science department parameters.

[0025] SBTA module: Spatio Bottom-up Top-down Module;

[0026...

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 behavior recognition system from bottom to top and from top to bottom. The behavior recognition system comprises an SBTA module and an STBTA module. The SBTA module and the STBTA module encode local features and global information through a bottom-to-top mechanism and an attention mechanism. The module of the invention can directly capture long-range dependence in an appropriate area in an image or a scene; channel statistics and spatial grid statistics are generated by using the maximum pool and the average pool; the sensitivity to an information function is improved, useful information is selected, the focusing position can be selected, and different representations of the position object can be enhanced; the method provided by the invention is a feedforward mode, and can be directly inserted into a 2D/3D CNN as an effective, simple and interpretable method; even if only STBA and STBTA exist, the performance is greatly improvded.

Description

technical field [0001] The invention relates to behavior recognition, in particular to a bottom-up-top-down behavior recognition system. Background technique [0002] Currently, human action recognition in videos occupies an important position in computer vision and has attracted a lot of attention. CNN based methods have achieved great progress in image classification. Furthermore, image classification tasks have more labeled images to train networks than labeled video data. In view of these two points, many methods combine predictions from images of videos by image-based classification methods to classify videos. However, videos not only possess many irrelevant information related to human actions in and between frames, but also include more temporal information along frames, i.e., long-range temporal dependencies. [0003] In vision tasks, some methods try to capture long-range dependencies. Some modules use separate backbones to independently process images at multip...

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): G06K9/00G06N3/04
Inventor 招继恩朱勇杰王国良张海谭大伦周明
Owner GUANGZHOU INTELLIGENT CITY DEV INST
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