Football Detection in Video using YOLOv8 and Deep SORT

This project demonstrates the use of YOLOv8 and Deep SORT for detecting and tracking a football in video footage. The custom data used for this project was annotated using Roboflow.

Table of Contents

• Introduction
• Features
• Usage
• Roboflow
• Results
• Contributing
• License

Introduction

In this project, we employ the YOLOv8 model for detecting a football in video frames and the Deep SORT algorithm for tracking the detected football across the frames. This combination ensures robust and real-time object detection and tracking.

Features

• YOLOv8: An advanced object detection model that is efficient and accurate.
• Deep SORT: An algorithm for multi-object tracking that combines Kalman filtering and Hungarian algorithm for data association.
• Custom Data: The dataset used for training the model was annotated using Roboflow.

Usage

• Clone the Repository:

  git clone https://github.com/cizodevahm/Soccer-Ball-Tracking-Machine-Learning.git

  • Run the Football__YOLOv8_Detection_Tracking_CustomData (2).ipynb file for training on custom data.
  • Change the model name if you want to experiment like yolov8m, yolov8s.

Roboflow

Roboflow is a platform that simplifies the process of collecting, annotating, and preparing image data for machine learning projects. With Roboflow, you can:

• Annotate Images: Easily annotate images using a user-friendly interface.
• Augment Data: Apply various augmentation techniques to enhance the dataset.
• Export Data: Export the annotated data in formats compatible with popular machine learning frameworks.

Results

• Here are some sample results from our project:

Contributing

• Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

• This project is licensed under the GPL-3.0 license. See the LICENSE file for more details.