Drowning Detection Application

A real-time drowning detection application using a YOLOv8-based deep learning model. This project processes video or image inputs to detect events related to drowning, swimming, or being out of water. When a drowning event is detected, the system draws bounding boxes around the detected area and plays an alert sound.

Table of Contents

• Overview
• Features
• Prerequisites
• Installation
• Usage
• Project Structure
• Model Training
• License

Overview

This project implements a real-time detection system that:

• Processes both images and videos to detect drowning events.
• Differentiates between drowning, swimming, and out of water events.
• Plays an alert sound (sound/alarm.wav) whenever a drowning event is detected.

The system leverages a YOLOv8 model that has been previously trained using a dataset (e.g., from Roboflow) containing images with the three classes. The application uses OpenCV for video processing and the playsound module for audio alerts.

Features

• Real-Time Detection: Process live video streams or input files to detect events.
• Multi-Class Support: Identify and display bounding boxes for drowning, swimming, and out of water events.
• Audible Alerts: An alarm sound is played when a drowning event is detected.
• User-Friendly: Simple command-line interface to specify the input source.

Prerequisites

• Python 3.7+
• Required Python packages:
  • opencv-python
  • ultralytics
  • playsound

Installation

1. Clone the Repository:

   git clone <repository_url>
   cd <repository_directory>

2. Install Dependencies:

   pip install opencv-python ultralytics playsound

3. Model and Sound Setup:

Ensure that trained YOLOv8 model file (best.pt) is placed in the same directory as app.py. Place the alert sound file (alarm.wav) inside a folder named sound (i.e., sound/alarm.wav).

Usage

Run the application by specifying the input file (image or video):

• For an image:

  python app.py --source path/to/your/image.jpg

• For a video:

  python app.py --source path/to/your/video.mp4

When executed, the application will display the input with annotated bounding boxes. If a drowning event is detected, an audible alarm will be played.

Project Structure

.
├── app.py             # Main application file
├── best.pt            # Trained YOLOv8 model weights
├── sound
│   └── alarm.wav      # Alert sound file
└── README.md          # Project documentation

Model Training

The YOLOv8 model used in this application was trained on a dataset from Roboflow containing images labeled as drowning, swimming, and out of water. You can see the training code in "Drown_detect.ipynb".

License

This project is licensed under the MIT License.