Skip to content

Latest commit

 

History

History
88 lines (55 loc) · 2.6 KB

README.md

File metadata and controls

88 lines (55 loc) · 2.6 KB

Sports Classification using CNN, Xception, and ResNet

This repository contains code for classifying sports images using Convolutional Neural Networks (CNN), Xception, and ResNet architectures. The project demonstrates how to preprocess image data, build and train deep learning models, and evaluate their performance.

Table of Contents

Introduction

The goal of this project is to classify images into different sports categories using deep learning. We explore and compare three architectures:

  • A custom Convolutional Neural Network (CNN)
  • Xception
  • ResNet (Residual Networks)

Installation

  1. Clone the repository:

    git clone https://github.com/your-username/sports-classification.git
cd sports-classification

  2. Create and activate a virtual environment:

    python3 -m venv venv
source venv/bin/activate  # On Windows use `venv\Scripts\activate`

  3. Install the required dependencies:

    pip install -r requirements.txt

Dataset

The dataset should be organized into train, val, and test directories, with subdirectories for each sport category. Each subdirectory contains the images belonging to that category. https://www.kaggle.com/datasets/gpiosenka/sports-classification/data

Usage

  1. Data Preprocessing: To preprocess the images and prepare the dataset.

  2. Model Training: We can choose between the CNN, Xception, and ResNet models by specifying the appropriate script from the models directory.

  3. Evaluation: This will generate metrics and visualizations to compare the performance of the different models.

Models

CNN

A custom CNN architecture designed for sports image classification.

  • Simple and fast to train

Xception

An implementation of the Xception architecture, known for its depthwise separable convolutions and high accuracy.

  • Requires more computational resources

ResNet

An implementation of the ResNet architecture, which introduces residual learning to ease the training of deep networks.

  • Robust and widely used in various image classification tasks

Results

The results of the trained models, including accuracy, precision, recall, and F1-score, are documented in the notebooks/evaluation.ipynb notebook. We can also find visualizations of the model performances and confusion matrices.

License

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