This repository contains data for COVID19 and a Transfer Learning Based model for classification
There are two types of classification that were done.
- A four class classification among X-Ray Images. The classes were Nomral X Rays, COVID 19 X Rays, Viral Pneumonia, Bacterial Pneumonia etc. Run four_class.py for this.
- A two class classification on COVID 19 X rays and Normal X Rays. Run two_class.py for this.
The COVID-19 X-ray image dataset was curated by Dr. Joseph Cohen, a postdoctoral fellow at the University of Montreal. If you want to contribute in his work, please visit his Github Repository . The other types of data samples were collected from kaggle and some other sources.
I used RESNET-50 based transfer learning for training. As the dataset is fairly very small I augmented the dataset. Furthermore, as the number of samples is fairly low, the accuracy is very low. To build a more powerful classification model, more data is needed.
Currently the dataset is divide evenly.
- Nummber of COVID Samples - 111 ( Train:80, Test: 31)
To balance the dataset, other class samples were taken according to the number of COVID Samples.
Requirements
- Tensorflow
- OpenCV
- Numpy
Below there are two samples of X Ray Images
| A sample of COVID-19 X Ray Image | A sample of Normal X Ray Image |
|---|---|
 |
 |