Skip to content

Description PoC for the 2020 WirVSVirus Hackathon

Notifications You must be signed in to change notification settings

adrigru/ai4covid

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AI4COVID

This project is a result of the WirVsVirus Hackathon. Our objective was to assist doctors in diagnosing COVID-19 patients by training a convolutional network to discriminate between patients with and without the disease, based on an X-ray image. We use a DenseNet121 pretrained on the CheXpert dataset and finetune it to the covid-chestxray-dataset. Furthermore, we implemented a simple prediction service that displays the diagnosis based on an X-ray image uploaded by the user.

alt text

How does it work?

  1. We first trained the COVID-19 classifier on Google Colab. See the notebook file covid_19.ipynb.
  2. We then implemented a simple backend server in Flask, that loads the classifier and performs inference on the X-ray image uploaded by the user.
  3. Finally, we developed a simple frontend server that allows user to upload an X-ray image and see the results of diagnosis: the predicted probability of patient having COVID-19 and the heatmap of critical regions on the X-ray that contributed to the diagnosis.

Setup

Requirements:

  • python >=3.7.6
  • NPM >=6.13.4

Running the application

  1. cd ai4covid -- Enter project root directory
  2. chmod +x setup.sh run.sh -- Add execute permission to the scripts
  3. ./setup.sh -- Install dependencies
  4. ./run.sh -- Run the backend and the frontend servers