Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

A training method and application of a head and shoulder image classifier

An image classifier and training method technology, which is applied in the fields of instrument, calculation, character and pattern recognition, etc., can solve the problem of low head and shoulders detection ability, improve the ability of head and shoulders detection, avoid poor adaptability, and achieve reliable recognition results. Effect

Active Publication Date: 2019-06-18
HUAZHONG UNIV OF SCI & TECH
View PDF7 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a training method and application of a head and shoulders image classifier to solve the problem of low head and shoulders detection ability existing 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
  • A training method and application of a head and shoulder image classifier
  • A training method and application of a head and shoulder image classifier
  • A training method and application of a head and shoulder image classifier

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0052] A training method 100 of a head and shoulders image classifier, such as figure 1 shown, including:

[0053] Step 110, obtaining a human body posture data set, which includes a plurality of human body sample pictures and human body head and shoulder annotation data corresponding to each human body sample picture;

[0054] Step 120, through the multi-core parallel algorithm and single instruction multiple data flow instruction set, synchronously calculate multiple human body sample pictures, and obtain the aggregation channel feature map corresponding to each human body sample picture and the multiple down-sampling corresponding to the aggregation channel feature map Aggregate channel feature maps;

[0055] Step 130: Determine the positive detection window group and its corresponding positive feature vector group and the negative detection window group and its corresponding negative feature vector group based on all aggregated channel feature maps and downsampled aggrega...

Embodiment 2

[0063] On the basis of Embodiment 1, the plurality of human body sample pictures include: human body sample pictures of various postures and multiple perspectives.

[0064] This embodiment aims at the problems that template matching and other methods based on artificial image features in the prior art are difficult to detect occluded human bodies, and it is difficult to adapt to changes in color and outline shape. A human head and shoulders data set containing multi-pose and multi-view human body images is used as training data , to avoid the poor adaptability of the detection method due to a single data source.

Embodiment 3

[0066] On the basis of Embodiment 1 or Embodiment 2, each aggregation channel feature map and each downsampling aggregation channel feature map include ten feature channels;

[0067] Then step 120 includes:

[0068] Through the multi-core parallel algorithm, multiple human body sample pictures are calculated synchronously, and the aggregation channel feature map corresponding to each human body sample picture and the multiple down-sampled aggregation channel feature maps corresponding to the aggregation channel feature map are obtained. When channel feature maps and each down-sampled aggregation channel feature map, the above ten feature channels are calculated using SIMD instruction set.

[0069] The multi-core parallel algorithm is used to calculate the feature map, and the single instruction multiple data stream instruction set is used to calculate the feature channel of the feature map, which greatly improves the speed of data processing in the case of a large amount of da...

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 relates to a training method and application of a head and shoulder image classifier, and the method comprises the steps: obtaining a human body posture data set which comprises a plurality of human body sample pictures and human body head and shoulder labeling data corresponding to each sample picture; Synchronously obtaining an aggregation channel feature map and a downsampling aggregation channel feature map corresponding to each human body sample picture through a parallel algorithm; Based on the feature map, determining a positive detection window group and a corresponding positive feature vector group, and a negative detection window group and a corresponding negative feature vector group; And on the basis of the positive feature vector group and the negative feature vector group, performing multi-stage training on the enhanced decision tree to obtain a head and shoulder image classifier. Through a parallel algorithm, the head and shoulder detection speed is improved, and head and shoulder detection can be carried out in real time in a video stream; A feature map containing a plurality of feature channels is used for training, and a classifier result obtained through training is more reliable; And multi-stage training is carried out on the enhanced decision tree, so that the identification precision of the head and shoulder image classifier is greatly improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a training method and application of a head and shoulders image classifier. Background technique [0002] Human detection is an important research topic in computer vision, and it is widely used in intelligent video surveillance, vehicle assisted driving, intelligent transportation, human-machine collaborative intelligent robots and other fields. The mainstream pedestrian detection methods are divided into statistical learning methods based on artificial image features and deep learning methods based on artificial neural networks. The logarithmic average missed detection rate of the statistical learning method for human objects is generally about 15%, which can be detected in real time on the CPU, but is easily affected by occlusion. The logarithmic average missed detection rate of the deep learning method is as low as 7%, but it needs to use GPU for calculation, which i...

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): G06K9/00G06K9/62
Inventor 彭刚陆泽早
Owner HUAZHONG UNIV OF SCI & TECH
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
Eureka Blog
Learn More
PatSnap group products