A Pedestrian Re-ID Method Based on Decoupled Adaptive Discriminative Feature Learning
A technology of pedestrian re-identification and feature learning, which is applied in the field of pedestrian re-identification based on decoupling adaptive discriminative feature learning, can solve the problem of insufficient optimization of pedestrian feature learning network, and achieve the effective effect of the method
Active Publication Date: 2021-08-10
ZHEJIANG UNIV
View PDF6 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0006] The invention provides a pedestrian re-identification method based on decoupling adaptive discriminative feature learning, which solves the problem of insufficient optimization of the current pedestrian feature learning network, and achieves a comparatively traditional network based on the existing model structure. Good effect of optimization method
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 in detail below with reference to the accompanying drawings and embodiments. It should be noted that the following embodiments are intended to facilitate the understanding of the present invention, but do not limit it in any way.
[0034] Such as figure 1 As shown, a pedestrian re-identification method based on decoupling adaptive discriminative feature learning includes the following steps:
[0035] S01, select an existing pedestrian re-identification model, the network before extracting the feature vector in the model is called the feature extraction layer, and the subsequent fully connected layer is called the classifier layer.
[0036] In this embodiment, the DenseNet161 network is taken as an example to illustrate each module. Input: a pedestrian photo, the input size is 256(height)*128(width).
[0037] Considering the importance of model initialization, this embodiment uses the DenseNet161 pre-trained by the ImageNet...
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
The invention discloses a pedestrian re-identification method based on decoupling adaptive discriminative feature learning, including: (1) selecting an existing pedestrian re-identification model, and dividing the model into a feature extraction layer and a classifier layer; (2) In the training phase, the parameters of the classifier layer are randomly initialized after each N passes of data training, the learning rate of the feature extraction layer decreases with the iteration of the data, and the learning rate of the classifier layer remains unchanged; training until the objective function converges; (3) In the test phase, only the feature extraction layer is kept as a trained network model; (4) In the pedestrian retrieval phase, the trained network model is used to extract the feature vector of each picture in the picture library, and the pedestrian pictures to be queried The feature vector and the feature vector of each picture in the picture library are sorted by similarity and the identity of the top-ranked picture is selected as the final recognition result. The present invention can well solve the problem of insufficient optimization of the current pedestrian feature learning network.
Description
technical field [0001] The invention belongs to the field of computer vision and pattern classification, in particular to a pedestrian re-identification method based on decoupling adaptive discriminative feature learning. Background technique [0002] In recent years, with the emergence of a large number of surveillance cameras in public places, pedestrian re-identification technology has received more and more attention. The goal of pedestrian re-identification technology is to perform cross-camera pedestrian search, that is, given a pedestrian picture captured by a certain camera, find pictures belonging to the same identity as this picture from other cameras. Pedestrian re-identification has a wide range of applications in missing person finding and suspect tracking. In recent years, pedestrian re-identification technology has developed rapidly, but it still faces many challenges in actual scenes, including: illumination changes under different cameras, human posture cha...
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 Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04
CPCG06V20/53G06N3/045G06F18/22G06F18/214G06F18/2415
Inventor 魏振勇魏龙蔡登金仲明黄建强华先胜何晓飞
Owner ZHEJIANG UNIV