Covid-Detection-from-Chest-X-Ray

This project deals with the covid-19 detection from chest X-ray using various methods and a comparitive study between them.

Colab File : https://colab.research.google.com/drive/1LWDPv8UWCeCEcH-9LFtXzwVO5HkJrEe6#scrollTo=wFKyrDpVd2DI

Resources :

• HOG FEATURES.
• OPEN CV
• HOG
• HOG BASICS
• LBP

This repository contains code for detecting COVID-19 from chest X-ray images using three different techniques: Histogram of Oriented Gradients (HOG), Convolutional Neural Network (CNN), and Local Binary Pattern (LBP).

Introduction

The outbreak of COVID-19 has posed a significant challenge worldwide, and the early and accurate detection of the disease is crucial for effective management and control. Chest X-ray imaging has emerged as a valuable tool for diagnosing COVID-19 due to its wide availability and rapid turnaround time. In this project, we explore three different techniques for automated COVID-19 detection from chest X-ray images.The quality of the chest-X-ray images is not good , so a lot of preprocessing is required .

Techniques Used

1. Histogram of Oriented Gradients (HOG):

   • HOG is a feature descriptor widely used in object detection and image classification tasks.
   • We extract HOG features from chest X-ray images and feed them into a machine learning model for COVID-19 detection.

2. Convolutional Neural Network (CNN):

   • CNNs are deep learning models known for their effectiveness in image classification tasks.
   • We train a CNN model on a dataset of chest X-ray images to learn features and classify them into COVID-19 positive or negative.

3. Local Binary Pattern (LBP):

   • LBP is a texture descriptor used for texture classification and face recognition.
   • We extract LBP features from chest X-ray images and use them to train a machine learning model for COVID-19 detection.

Repository Structure

• implementation: Contains from scratch implementations of lbp and hog.
• model: Contains different model implementation.
• Covid_Detection_Using_X_Ray.ipynb/: Contains Python scripts for implementing HOG, CNN, and LBP techniques for COVID-19 detection.
• README.md: This file, providing an overview of the project.

Usage

1. Clone the repository:

   git clone https://github.com/sahaniaditya/Covid-Detection-from-Chest-X-Ray.git
   git clone https://github.com/shikhar5647/Covid-Detection-from-Chest-X-Ray.git

2. Navigate to the project directory:

   cd Covid-Detection-from-Chest-X-Ray

3. Follow the instructions in the respective directories (code/) to run the code for each technique.

Results

• HOG feature extraction technique and the implementation of various ML models achieved the best accuracy of 74.56%.
• CNN technique achieved an accuracy of 91.66%. Both Tensorflow and Pytorch implementation were done and the results demonstrated.
• LBP technique achieved an accuracy of 95.66%. A simple neural network was implemented after the extraction of the features.

Conclusion

In this project, we explored three different techniques for COVID-19 detection from chest X-ray images. Each technique has its advantages and limitations. Further research and experimentation could lead to improved models for more accurate and reliable COVID-19 detection.

Contributors

• Aditya Sahani
• Shikhar Dave