Image-classification-projects

This repository contains image classification tasks:

• digit recognition classification.
• rock-paper-scissors classification.
• Intel image classification

The models are implemented using deep learning techniques and are trained on suitable datasets to achieve high accuracy.

---

Introduction

The purpose of this repository is to provide pre-trained models and code for image classification tasks, specifically digit recognition and rock-paper-scissors classification. These models can be used for various applications, such as automatic digit recognition in handwritten documents or playing rock-paper-scissors with a computer.

---

Installation

To use the models and code in this repository, follow these steps:

1. Clone the repository to your local machine:

   git clone https://github.com/tawfikhammaf/Image-classification-projects.git
   
   

2. Install the required dependencies. Assuming you have Python and pip installed, run the following command:

   pip install -r requirements.txt
   
   

3. Setup the environment by downloading the necessary datasets and pre-trained models:

   python setup.py
   
   
   This command will download the required datasets and pre-trained models into the appropriate directories.
   

---

Models

This repository currently provides the following models:

1. Digit Recognizer: This model is trained to recognize and classify handwritten digits from 0 to 9.

2. Rock-Paper-Scissors Classifier: This model is trained to classify images of human hand gestures representing rock, paper, or scissors.

3. Intel images classification: This model is trained to classify images of Natural Scenes around the world distributed under 6 categories.

---

Dataset

The models in this repository were trained on the following datasets:

• MNIST: The MNIST dataset consists of 60,000 training images and 10,000 test images of handwritten digits. It is widely used for digit recognition tasks.

• Rock-Paper-Scissors: The rock-paper-scissors dataset consists of images of hands displaying rock, paper, or scissors gestures. It contains a total of X images, divided into training and test sets.

• Intel images: This is image data of Natural Scenes around the world that contains around 25k images of size 150x150 distributed under 6 categories.

The datasets used for training and evaluation are available for download from the following sources:

• MNIST: https://www.kaggle.com/competitions/digit-recognizer/data
• Rock-Paper-Scissors: https://www.kaggle.com/datasets/frtgnn/rock-paper-scissor
• Intel image classification: https://www.kaggle.com/datasets/puneet6060/intel-image-classification

---

Results

The accuracy and performance of the models are as follows:

1. Digit Recognizer: Achieves an accuracy above 99.38% on the MNIST test dataset.

2. Rock-Paper-Scissors Classifier: Achieves an accuracy of 98.3% on the rock-paper-scissors test dataset.

3. Intel images classification: Achieves an accuracy of 91.3% on the test dataset.

License

The code and models in this repository are available under the MIT License. Feel free to use them for academic, research, or commercial purposes.