Target detection method and device based on multi-gating hybrid expert model

A hybrid expert and target detection technology, applied in neural learning methods, biological neural network models, character and pattern recognition, etc., can solve problems such as damage to target detection accuracy, rough joint learning strategies, etc., to improve detection accuracy, avoid The effect of negatively affecting each other and improving efficiency

Active Publication Date: 2022-03-01
BEIJING KITTEN & PUPPY TECH CO LTD
View PDF14 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] However, the joint learning strategy of object classification and bounding box regression is still relatively rough. Usually, in the prior art, only the description of the above two tasks is shared, but simply sharing the parameters of the two tasks may damage the accuracy of target detection.

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
  • Target detection method and device based on multi-gating hybrid expert model
  • Target detection method and device based on multi-gating hybrid expert model
  • Target detection method and device based on multi-gating hybrid expert model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0071] Embodiment 1 provides an object detection method based on a multi-gated mixed expert model, which solves the problem of multi-task joint learning of object classification and bounding box regression, thereby improving the accuracy of the object detection method. The overall framework is as figure 2 As shown, first input the image into the deep convolutional neural network to obtain multi-scale features, then input the multi-scale features into the feature pyramid network to obtain the fused multi-scale features, further, input the multi-scale features into the The area recommendation network and the area pooling layer process to obtain the candidate area features, and finally input the candidate area features into the multi-gated mixed expert model to obtain the category and location of the object. Its process is as follows figure 1 Shown, specifically, described method comprises the following steps:

[0072] S1: Obtain the target feature map of the region where the ...

Embodiment 2

[0135] Image 6 It is a schematic structural diagram of a target detection device based on a multi-gated mixed expert model provided according to a specific embodiment of this patent. like Image 6 As shown, the system includes: an acquisition module 60 , a first processing module 61 , a second processing module 62 , and a determination module 63 .

[0136] Obtaining module 60, acquiring the target feature map and the potential target frame of the region where the potential target is located in the image;

[0137] The first processing module 61 processes the target feature map using an expert model, and outputs the target classification subtask result and the bounding box regression parameter subtask result corresponding to the target feature map; the quantity of the expert model includes multiple Each expert model outputs its target classification subtask results and bounding box regression parameter subtask results respectively;

[0138] The second processing module 62 pr...

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 target detection method and device based on a multi-gating hybrid expert model. The method comprises the steps of obtaining a target feature map and a potential target frame of an area where a potential target is located in an image; processing the target feature map by using an expert model, and outputting a target classification subtask result corresponding to the target feature map and a frame regression parameter determination subtask result; processing the target feature map by using a gating network, and outputting an adaptive weight value of each expert model corresponding to the target classification sub-task and an adaptive weight value of each expert model corresponding to the frame regression parameter determination sub-task; and according to the adaptive weight value, the target classification subtask result and the frame regression parameter subtask result, determining the category and the frame of the target through full-connection neural network processing. And target classification and regression learning are performed through the multi-gating hybrid expert model, so that the efficiency of classification and regression task joint learning is improved, and the accuracy of target detection is improved.

Description

technical field [0001] This patent relates to the field of computer vision target detection. Specifically, it relates to a target detection method and device based on a multi-gated mixed expert model. Background technique [0002] In recent years, deep learning technology has developed rapidly, and the field of computer vision has ushered in an era of rapid development. The research and application of computer vision algorithms in academia and industry are emerging in an endless stream, with a wide range of landing scenarios, which have had a huge impact on human life. [0003] As an aspect of the field of computer vision, object detection is also dominated by deep learning methods. Object detection refers to predicting the category and location of all possible objects on the image. The basic process is to extract the depth features of the image through the deep convolutional neural network, and then predict the areas where objects may exist through the area recommendation...

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): G06V10/764G06V10/80G06V10/22G06V10/766G06V10/82G06N3/04G06N3/08
CPCG06N3/08G06N3/084G06N3/045G06F18/2415G06F18/253
Inventor 吴琎何振东
Owner BEIJING KITTEN & PUPPY TECH CO LTD
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