ymzhu99 / traffic_flow_counter

An end-to-end implementation of a computer vision model for tracking vehicles from video.

Traffic flow counter 🚦

Introduction

Hello everyone! The following code will be a current work-in-progress app for traffic flow counting. My hope is to make this something of value for my city to be used for traffic flow management.This uses the yolo-v3 computer vision model to vehicles.

Setup

To run the app locally, install the necessary python packages by:

pip install -r requirements.txt

For better reproducibility, make sure to open up a python environment using virtualenv or any of your favorite python environment packages.

The app contains a custom slider. To use it, make sure you have npm installed and then run the following commands.

cd components/custom_slider/frontend/
npm install 

Afterwards, run npm run build

This should recreate the build package for the custom slider.

pull the necessary models and data by running:

make dependencies 

Run the app!

streamlit run app/streamlit-app.py

Future Developments

Here are some of my ideas that would be neat additions to the API:

• Speed measurement
• Vehicle distribution measurement
• Traffic density