COVID19_Detection_Project

Authors: Ji Hoon Chung, Kibae Kim

Overview

Using data pulled out from Kaggle, we retrieved chest X-Ray of normal people & COVID-19 patients:
Data Imported from = https://www.kaggle.com/donjon00/covid19-detection
*NORMAL IMAGES = 11,767 Images
*COVID-19 IMAGES = 3,616 Images

---

Coronavirus disease 2019 (COVID-19) is a contagious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The disease has since spread worldwide, leading to an ongoing pandemic. So, we built a deep learning model to detect people who infected to COVID-19 virus by using their lung X-ray data from Kaggle. This project will help hospitals figure out infected people when they don’t have COVID test kit.

Business Problem

---

Hospitals are short with COVID-19 test kits, and they are looking for alternative ways to detect COVID-19 infection status. We are looking to build a model which can detect COVID-19 just by looking at Chest X-ray images.

Modeling

---

We've tried 3 different models to build COVID-19 classification project.

1. Densely Connected Network Model
   
   
   

2. Baseline CNN Model
   
   
   

3. Weighted CNN Model
   
   
   

Conclusion

---

1. We chose weighted CNN model because this model detect COVID-19 well and also detect normal well.
2. Our weighted CNN model can be useful as a method of COVID-19 detection for hospitals when they don’t have test kit.

Ideas for Improvement

---

1. See if the model can differentiate COVID X-ray images from other lung disease X-ray images such as pneumonia.
2. See if we can develop new models to detect other diseases using X-ray images.

Repository Structure

---

├── README.md                                      <- The high-level overview of this project
├── COVID19_Detection_project_presentation.pdf     <- PDF version of project presentation
├── COVID19_Detection_Project.ipynb                <- Final_Notebook used for the project
├── images                                         <- Sourced externally and visualizations generated from code
├── data                                           <- All the Image data files used for the notebook.