A method for synergistically training a face recognition network and a pedestrian re-recognition network
A pedestrian re-identification and face recognition technology, applied in the field of deep learning, can solve the problems of inefficient pedestrian re-identification, ignoring the correlation between face recognition and pedestrian re-identification, etc., to achieve strong facial feature expression ability and strengthen influence , improve the effect of accuracy
Active Publication Date: 2019-01-11
CHINA JILIANG UNIV
View PDF8 Cites 6 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, face recognition and pedestrian re-identification are usually studied as two independent tasks, ignoring the correlation between face recognition and pedestrian re-identification, and relying only on information such as clothes and height for pedestrian re-identification is not efficient.
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0033] The present invention will be further described below in conjunction with accompanying drawing.
[0034] In this example, if figure 1 As shown, a flow chart of collaborative training of a face recognition network and a pedestrian re-identification network. The specific implementation mainly includes the following steps:
[0035] Step (1): Use the face detection module of the open source face recognition engine SeetaFace to perform face detection on the DukeMTMC-reID pedestrian re-identification dataset. The pedestrian dataset uses the DukeMTMC-reID pedestrian re-identification database, including 702 pedestrians There are 16,522 images in , with an average of 23.5 training data for each type of pedestrian. The face detection module adopts a funnel-structured cascade structure (Funnel-Structured Cascade, FuSt), and the FuSt cascade structure is composed of a plurality of fast LAB cascade classifiers for different postures at the top, followed by several SURF-based featu...
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
Login to View More Abstract
A method for synergistically training a face recognition network and a pedestrian re-recognition network is disclosed. Based on the dual-network parallel network structure, face and pedestrian featurefusion is carried out, the fused features are taken as pedestrian output features, so that higher facial feature expression ability is exhibited, and according to the different definition of face image, the face recognition network and pedestrian re-recognition network are trained with different supervisory signals. When the definition of face image is low, the predicted result of pedestrian re-recognition network and the weighted result of real label are used as supervisory signals to guide the training of face recognition network. When the face image is clear, the predicted result of face recognition network and the weighted result of real label are used as supervisory signals to guide the pedestrian rerecognition network training, which not only enhances the influence of face recognition on pedestrian rerecognition results, but also can use the predicted result of pedestrian rerecognition to guide the classification of face features when the face image is blurred.
Description
technical field [0001] The invention belongs to the field of deep learning for extracting facial features by a deep neural network, relates to neural network, pattern recognition and other technologies, and in particular relates to a collaborative training method of a face recognition network and a pedestrian re-identification network. Background technique [0002] With the rapid development of my country's safe city construction, it is very important for security and public security criminal investigation business to use many video surveillance cameras to quickly and accurately obtain portrait information in the scene. Fast and effective automatic identity verification is becoming more and more urgent in the field of security. The use of video surveillance for face recognition and pedestrian re-identification has attracted more and more attention from the public security department. [0003] The current face recognition and pedestrian re-identification technology achievemen...
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
Login to View More IPC IPC(8): G06K9/00
CPCG06V40/161G06V40/168G06V40/172
Inventor 章东平陶禹诺陈思瑶毕崇圆郑寅
Owner CHINA JILIANG UNIV