Cross-Camera-Multi-Person-Tracking

Introduction

• Implementation by Wei-Cheng Lin, Service Systems Technology Center, Industrial Technology Research Institute, Taiwan.
• Develop a client-server person re-identification system to track people under 4 ipcams in a single room. In our setting, we have overall 3 rooms and 12 ipcams.

Operating System

• Windows 10
• Ubuntu 18.04

Install

conda create -n reid python=3.7
conda activate reid
pip install -r requirements.txt

Test under Pytorch 1.7 & Cuda 11.0, please intall them from their offical website.

Clone the TorchReid repository and build it from source.

git clone https://github.com/KaiyangZhou/deep-person-reid.git
cd deep-person-reid
python setup.py develop

Configuration

Download the model weights (you prefer) and put them into weight folder.

yolov7.pt yolov7x.pt yolov7-w6.pt yolov7-e6.pt yolov7-d6.pt yolov7-w6-person.pt

osnet_ain_x1_0 osnet_x1_0

Rename the reid model weight into osnet_ain_x1_0.pth and osnet_x1_0.pth, respectively

Head to config.py file, and modify the ipcam's address and additional setup as you want.

We use four ipcams as our default setup.

Inference

• Step 1 : In the client computer, run

python client.py

• Step 2 : In the server computer, run

python server.py

Be sure to run the client.py file before runing the server.py file.

Acknowledgement

Big thanks to Multi-Camera-Multi-Person-Tracking for sharing code.

• Multi-Camera-Multi-Person-Tracking
• TorchReid
• Yolov7