Unlock instant, AI-driven research and patent intelligence for your innovation.

A system for adjusting sitting posture of person in vehicle according to recognized person information

An adjustment system and in-vehicle technology, applied in the field of facial recognition, can solve the problems of inconvenience of adjustment without fundamental change, inability to automatically complete adjustment, limited number of personnel, etc., to achieve reduced differences, sufficient feature expression, and improved robustness Effect

Active Publication Date: 2019-08-16
MOMENTA SUZHOU TECH CO LTD
View PDF5 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although this method greatly simplifies the adjustment steps, it also has disadvantages: 1. The number of personnel that can be set is limited, and often only a few groups of personnel information can be memorized; when there are other personnel on the vehicle who have not set parameter groups, they cannot Automatically complete the adjustment; 2) It still requires personnel to select the parameter group option that suits them, so the inconvenience of adjustment has not been fundamentally changed
[0004] like figure 1 As shown, in the traditional method, each graphics card shares the same data input and the same classifier; generally in the industry, a commercially available face recognition model is often based on a training set of millions or even tens of millions of people. However, generally on a graphics card with a single video memory of 12G, when the dimension of the feature vector is 512 dimensions, the number of classifiers is generally at most about 300,000, otherwise it will face problems such as video memory overflow
Large training sets and video memory limitations have become the main contradiction in face recognition model training

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 system for adjusting sitting posture of person in vehicle according to recognized person information
  • A system for adjusting sitting posture of person in vehicle according to recognized person information
  • A system for adjusting sitting posture of person in vehicle according to recognized person information

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 2

[0038] The present embodiment provides a face recognition system, which specifically includes a face recognition model, which is obtained through deep learning training. The network model is composed of a data input layer, a feature fusion layer, a classifier and a loss function, wherein the loss function is a Softmax function .

[0039] The data input layer continuously traverses the training sample images; the feature fusion layer extracts the depth features of each picture;

[0040] The classifier classifies the sample image; the loss function compares the classification result with the real label of the sample image.

[0041] The above system also includes a detection device, which is used to detect the number and size of graphics cards in the system. According to the number and size of graphics cards, the data in the training set is allocated to the corresponding number of data for each graphics card; the specific allocation rules are as follows: if it is detected that th...

Embodiment 3

[0057] This embodiment provides a training method for a face recognition model, and the method is implemented by a face recognition training system.

[0058] Step S1: The detection device of the face recognition training system detects the number and size of graphics cards in the system, and inputs a corresponding amount of data in the training set to each graphics card according to the number of graphics cards and the size of the video memory. When the number of training set categories divided by the number of graphics cards is less than the maximum number of classifications of the classifier, the input data between each graphics card will generally overlap. For example, suppose there are 8 graphics cards, and the training set has 800,000 categories. The so-called overlap means that after ensuring that the classifier of each card is divided into 100,000 categories on average, 200,000 categories are randomly selected from the other 7 cards. Add a category to the classifier of ...

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

A system for adjusting the sitting posture of a person in a vehicle according to recognized person information comprises a sensor activation unit, a storage unit, an authentication system unit, a seatadjustment unit and the like. The sensor activation unit is used for detecting whether a person is located on a seat or not in the vehicle; the memory unit is used for storing a plurality of pieces of pre-stored vehicle personnel information; the authentication system unit comprises a neural network unit. In the prior art, for training of a neural network, a large training set and video memory limitation become the main contradiction in face recognition model training. The invention provides a training system of a face recognition model. The training system comprises a data input layer, a feature fusion layer, a classifier and a loss function. Different classifiers classify different contents of the sample image, so that the problem that a large-scale face data set cannot be trained due to video memory limitation in a training face recognition model is effectively solved.

Description

technical field [0001] The invention relates to the field of face recognition, in particular to the face recognition of a person in a vehicle driving environment, and the person's sitting posture in the vehicle is adjusted through the recognized person information. Background technique [0002] Face recognition: It is a biometric technology for identification based on human facial feature information. It is widely used in all aspects of human life. In the field of vehicle driving, because of reasons such as each person's gender, height, weight, etc., not a kind of seat adjustment position for driving or riding is applicable to everyone. Therefore people can only adjust the seat posture by some devices such as buttons arranged inside the vehicle, including the seat, the height of the back cushion, the front and rear positions of the seat, the inclination angle of the back cushion, etc.; In order to be friendly to human-computer interaction, the adjustment does not require m...

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): G06K9/62G06K9/00G06N3/08B60R16/02B60N2/02
CPCG06N3/084B60N2/02B60R16/02G06V40/16G06F18/24G06F18/214
Inventor 郑弘晖胡杰李雯雯
Owner MOMENTA SUZHOU TECH CO LTD