Object Detection on Live Georgia Traffic Camera Feeds

This is a proof-of-concept for our final project for CS 6476: Computer Vision - Spring '24 at Georgia Tech. It is a web application that uses object detection to detect vehicles in live traffic camera feeds from 511GA/Georgia Department of Transportation (GDOT). The application is built using Flask, YOLOv8, PyAv and PIL in the backend and Vanilla TypeScript and Mapbox GL JS in the frontend. For background work (fine-tuned models), check out this repository.

cover.mp4

Installation

Server

cd server
python3 -m venv venv
source venv/bin/activate # for windows: venv\Scripts\activate
pip install -r requirements.txt

Client

cd client
npm install

Running the Application

Server

Create a .env file in the server directory with the following contents:

GDOT_API_KEY=<your-api-key>

and then run the following commands:

cd server
source venv/bin/activate # for windows: venv\Scripts\activate
flask run # --debug for auto-reload

Client

Create a .env file in the client directory with the following contents:

VITE_MAPBOX_KEY=<your-mapbox-key>

and then run the following commands:

cd client
npm run dev