Overview / Usage

Person re-identification(re-ID) aims at finding the target person in different cameras, which is generally considered as an image retrieval problem. Different from the face recognition, the whole body image is employed in the identity process instead of the face data. It can be widely used in monitoring scenarios in daily life. My idea is to build a pedestrian retrieval system based on person re-ID technology to improve the safety of public places.[link:https://www.bilibili.com/video/av86739781]

​

​

Methodology / Approach

We use deep learning as a solution to the person re-ID task. Specifically, our method is as follows:

1. We designed a model structure that can learn useful pedestrian features from existing datasets.
2. In order to solve the problem of domain bias between the training dataset and real scenes, we propose a new domain adaptation method.
3. We used PyQT to write the interface and deployed the algorithm on NCS2 using the OpenVino tool.

Technologies Used

Python，Pytroch，OpenVino，HDDL， NCS2 ， PyQT，YOLO V3

Other links

• Video