Intelligent Traffic Light Vehicle Counting System

This repository contains the code for an intelligent real-time vehicle counting system at a traffic light.

• Videos used: Friantroulette YouTube Channel

Directory Structure

├── data/                  
│   ├── __init__.py
│   └── data_manager.py    
├── gui/                   # Graphical interface
│   ├── __init__.py
│   ├── app.py             
│   └── gui.py             
├── track/                 # Vehicle tracking
│   ├── __init__.py
│   └── tracker.py         
├── models/                # Trained model weights
│   └── model_main.pt      # <-- Place the model weights here
├── reports/               # Outputs and reports
│   ├── report.csv         # Example report in CSV format
│   └── report.png         # Example report image
├── main.py                # Application entry point
└── README.md              

Prerequisites

• Python 3.7 or higher
• pip

Install dependencies:

pip install -r requirements.txt

Download the Model Weights

Before running the system, download the trained model weights from the following link:

Download trained model

The file should be placed in models/model_main.py.

Usage

Run the complete application (GUI + counting):

python main.py

Initial Operation

1. In the interface, click Load video and select the desired video.
2. After the video loads, click two distinct points on the video preview area to define the counting baseline.
3. Click Play and the system will begin processing frames and display the real-time count in the interface.

Interface Components and Logic

1. data_manager.py: loads and pre-processes video frames
2. tracker.py: performs object-by-object vehicle tracking
3. gui.py / app.py: presents the real-time interface with updated counts
4. main.py: integrates all modules and starts the application

Contact

For questions, contact: jglg@cin.ufpe.br