AhmedSouley01/traffic-monitoring

🚦 Traffic Monitoring System

<div align="center"> ...

Real-time road traffic object detection, tracking and counting system

Developed as part of the Computer Vision course — AIMS Senegal — April 2026

Demo • Installation • Usage • Structure • Results

</div>

📋 Table of Contents

• —Description
• —Features
• —Demo
• —Architecture
• —Installation
• —Usage
• —Project Structure
• —Model
• —Detected Classes
• —Log Format
• —Results
• —Authors
• —License

📌 Description

This project implements a computer vision system dedicated to real-time road traffic monitoring. It is capable of:

• —Detecting vehicles and pedestrians in traffic videos
• —Tracking each object uniquely using ByteTrack
• —Counting the number of unique objects passing through the scene
• —Visualizing results via an interactive web interface
• —Analyzing data through a statistical dashboard

This system fits into a Senegalese national context where transport authorities are looking for reliable and automated methods to understand how traffic evolves across different regions and times of day.

✨ Features

• —✅ Real-time detection with fine-tuned YOLOv11
• —✅ Persistent tracking with ByteTrack (unique ID per object)
• —✅ Unique object counting (not just per-frame counting)
• —✅ Support for local video upload and online URL
• —✅ Class selection directly from the interface
• —✅ Color-coded bounding boxes per class with ID and confidence score
• —✅ Visual alert when no object is detected in the scene
• —✅ Automatic logs in CSV and JSON format
• —✅ Interactive dashboard with charts and statistics
• —✅ CSV export of filtered detection data
• —✅ Frame-by-frame processing with timestamps

🎬 Demo

📸 Screenshots available in the `results/images/` folder

🏗️ Architecture

Video (local or URL)
│
▼
┌───────────────┐
│ Pre-processing│  OpenCV — frame-by-frame reading
└───────┬───────┘
│
▼
┌───────────────┐
│   Detection   │  YOLOv11 — object detection
│   (YOLOv11)   │  conf=0.3, iou=0.45
└───────┬───────┘
│
▼
┌───────────────┐
│   Tracking    │  ByteTrack — unique and persistent
│  (ByteTrack)  │  ID per object across frames
└───────┬───────┘
│
▼
┌───────────────┐
│   Counting    │  ObjectCounter — unique object
│               │  counting per class
└───────┬───────┘
│
▼
┌───────────────┐
│   Logging     │  CSV + JSON — timestamps,
│               │  classes, bbox, track IDs
└───────┬───────┘
│
▼
┌───────────────┐
│  Web Interface│  Flask — annotated video stream
│               │  + statistical dashboard
└───────────────┘

⚙️ Installation

Prerequisites

• —Python 3.10+
• —pip
• —GPU recommended (CUDA 11.8+) — also works on CPU

Steps

1. Clone the repository

git clone https://github.com/your-username/traffic-monitoring-project.git
cd traffic-monitoring-project

2. Create a virtual environment

python -m venv venv

# Linux / macOS
source venv/bin/activate

# Windows
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Place the trained model

# Copy your fine-tuned model to:
models/traffic_yolo11_best.pt

If you do not have a fine-tuned model, the pre-trained YOLOv11 model will be downloaded automatically.

5. Launch the application

python app.py

6. Open in your browser

http://127.0.0.1:5000
http://192.168.1.41:5000

🚀 Usage

Via the web interface

1. 1.Open http://192.168.1.41:5000
2. 2.Choose a video source:
3. 3.Upload: select a .mp4, .avi, or .mov file
4. 4.URL: paste a link to an online video
5. 5.Select the classes to detect
6. 6.Enter a scene identifier (e.g. dakar_intersection)
7. 7.Click ▶ Start Detection
8. 8.View real-time results at /live
9. 9.Click ⏹ Stop to save the logs
10. 10.Open the Dashboard for analysis

Available arguments

📁 Project Structure

traffic-monitoring-project/
│
├── 📂 data/
│   ├── raw_videos/          # Original raw videos
│   ├── processed_videos/    # Annotated output videos
│   ├── annotations/         # Annotations for fine-tuning
│   └── schema.json          # Shared log schema
│
├── 📂 models/
│    └── traffic_yolo11_best.pt   # Fine-tuned model
│
├── 📂 src/
│   ├── detector.py          # YOLOv11 detection
│   ├── tracker.py           # ByteTrack tracking
│   ├── counter.py           # Unique object counting
│   ├── logger.py            # Log generation
│ 
│   ├── templates/
│   │   ├── base.html
│   │   ├── index.html       # Home page
│   │   ├── live.html        # Real-time detection
│   │   └── dashboard.html   # Statistics
│   └── static/
│       ├── css/style.css
│       ├── js/main.js
│
├── 📄 app.py                # Flask application
├── 📄 requirements.txt      # Python dependencies
├── 📄 README.md             # This file
├── 📄 LICENSE               # MIT License
└── 📄 report.pdf            # Final report

🤖 Model

YOLOv11 — Fine-tuning

Performance Metrics

Metrics will be updated after final training.

🎯 Detected Classes

📊 Log Format

CSV (logs/detections_*.csv)

scene_id, frame, timestamp, track_id, class, confidence, x1, y1, x2, y2
scene_01, 142,  00:04.73,  7,        car,   0.912,      120,340,280,420

JSON (logs/tracking_*.json)

{
  "scene_id":  "scene_01",
  "frame":     142,
  "timestamp": "00:04.73",
  "detections": [
    {
      "track_id":   7,
      "class":      "car",
      "confidence": 0.912,
      "bbox":       [120, 340, 280, 420]
    }
  ]
}

📈 Results

Analyzed Scenes

Results will be filled in after testing on the selected videos.

🛠️ Dependencies

flask>=3.0.0
ultralytics>=8.3.0
opencv-python>=4.8.0
numpy>=1.24.0
werkzeug>=3.0.0
PyYAML>=6.0

👥 Authors

Supervisor: Jordan F. Masakuna — AIMS Senegal

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

MIT License — Copyright (c) 2026 — AIMS Senegal

🙏 Acknowledgements

• —Ultralytics for YOLOv11
• —Kaggle for the datasets
• —Pexels for the traffic videos
• —AIMS Senegal for supervision and support

<div align="center"> <i>Built with ❤️ at AIMS Senegal — Computer Vision 2026</i> </div>