Air-Hockey-Using-CV

This repository contains the code for a computer vision-based air hockey game developed using OpenCV and Python. The game is designed to be played using a webcam or other camera device, with the computer vision algorithm tracking the movement of the puck and paddles in real time. The goal is to provide a virtual air hockey experience by leveraging object detection techniques.

Model Used

We have used a custom YOLOv8 model for circle detection to track the puck and paddles effectively during the game. The YOLOv8 model is chosen for its high speed and accuracy in detecting objects in real-time scenarios, making it ideal for a fast-paced game like air hockey.

Dataset

A custom dataset was created for this project, specifically designed for detecting circular objects like pucks. The dataset was annotated using the CVAT (Computer Vision Annotation Tool) platform, which allows for efficient and precise labeling of objects. The dataset includes images of pucks and paddles from various angles and lighting conditions to ensure robust detection.

Features

• Real-Time Object Detection: The game uses YOLOv8 to detect the puck and paddles in real-time, ensuring smooth gameplay.
• Custom Dataset: A dataset specifically built for detecting pucks and paddles, annotated using CVAT.
• OpenCV Integration: OpenCV is used for video capture, image processing, and displaying the game interface.
• Camera-Based Control: Players control their paddles using real-world objects tracked by the camera, adding an interactive layer to the game.
• Collision Detection: Implements basic physics for collision detection between the puck, paddles, and virtual table borders.

Installation

Our Work Environment

• CPU: Ryzen 5 5000 series
• RAM: 16 GB
• GPU: RTX 3050
• OS: Windows 11
• Python: 3.12.0

Setup Guide

1. Clone the repository:

   git clone https://github.com/Dhanay-J/Air-Hockey-Using-CV.git
   cd Air-Hockey-Using-CV
   

2. (Optional but recommended) Set up a Python virtual environment:

   python -m venv env
   

3. Activate the environment:

   • For Windows:

     .\env\Scripts\activate
     

   • For Linux/MacOS:

     source env/bin/activate
     

4. Install the required dependencies:

   pip install -r requirements.txt
   

Alternatively, you can manually install the libraries:

pip install opencv-python
pip install ultralytics
pip install pygame

Running the Game

To start the game, simply run:

python main.py

Make sure your webcam is connected and properly set up to detect the paddles and puck.

Using Your Own Models

Training and Testing Your Model

1. Create your own dataset:

   • You can collect and annotate your dataset using the CVAT or another annotation tool.

2. Train the model:

   • Use your custom dataset to train the YOLOv8 model. Adjust training parameters in config.yaml to match your dataset and hardware.

Checking if the Detection Works

1. After training your model, use the provided script yolo_test_run_model.py to check if the model detects the objects (puck and paddles) correctly.

   • This script is used for a basic functionality test, ensuring the detection works as expected. It does not calculate any metrics like accuracy or precision.

2. To use the model in the game, pass the model's path as a parameter to the Game object:

   game = Game(model_path="path/to/your/model.pt")
   

Replace "path/to/your/model.pt" with the actual path to your custom model.