sample readme.md

Table of contents

• Introduction
• Prerequires
• Installation
• How to use
• Todo

Branch note

• This version is added yolov5 module (detection: yolov5 onnx
• Model store at tienduchoangtb3@gmail.com

INTRODUCTION

This repo is first version of vehicle counting.

This repo is used to count vehicle include 2 modules:

• yolov4 for detect vehicle
• deep_sort for track vehicle

Methodology:

• Deploy: tensorrt inference server
• Communicate: Paho-MQTT
• Dockerize: docker-compose

PREREQUIRES

• Ubuntu 18.04
• Python3.7 + miniconda
• Nvidia-docker-compose version 2.3
• Trtis:20.02-py3
• Paho-mqtt 1.5.0

INSTALLATION

• Folow this instruction to install trtis:20.02-py3
• Pull model:

- Step 1: Install [dvc](https://dvc.org/doc/start) first
- Step 2: dvc pull
- Step 3: mail to tienduchoangtb@gmail.com to access my models

• Install the remaining component using pip

HOW TO USE

• Start service: docker-compose up
• To publish message from client to server: python3 pub.py
• See monitoring system at 0.0.0.0:3000

TODO

• Draw cross line
• Add counting algorithm
• Add threading
• Add MQTT
• Add logs
• Add docker-compose
• Notify in slack