COVID-19 Detection

Overview

The project is live here check it out.

This project involves End-to-end ML-project. from developing a Convolutional Neural Network (CNN) model based on the ResNet50 architecture to detect COVID-19 from X-ray images to building the API with the help if FastAPI. with containerinzation using Docker

The model was trained on a dataset with 148 examples, with F1 score 0.8696 and achieved an accuracy of 97.5%.

Requirements

1. GitHub
2. Docker

Files

• code.ipynb: Contains all the code for data preprocessing, model training, and evaluation.

• COVID DataSet.zip: The dataset used for training and testing the model.

• Dockerfile with the containerization of the project.

• The /app directory have all the file related to the Ops

  • With /app/main.py as the file containing FastAPI.
  • With /app/model.h5 the trained model.
  • With /app/requirement.txt having the required libraries.

Model Performance

• Accuracy: 97.5%
• F1 Score: 0.8696
• Training Data: 148 X-ray images

Usage

• Using deployed API:

  1. Just download the app/app.py file
  2. Run the following command in the terminal

     pip install streamlit
     streamlit run app.py

     by runing these commands a

         You can now view your Streamlit app in your browser.
     
         Local URL: http://localhost:8501
         Network URL: http://172.25.97.164:8501

     this type of window will be shown and you can open the app by following the link provided.

• Using Docker Image :

  1. Clone the Repository: Clone this repository
  2. Open the Repository: Go to the cloned repository
  3. Run Docker Image: use $ docker run covid-19-detection in the bash.
  4. Check the FastAPI: Check the API at the /predict/docs

• Manually :

  1. Clone the Repository: Clone this repository
  2. Open the Repository: Go to the cloned repository
  3. Go to the app:use $ cd app in the bash.
  4. Install required libraries run $ pip install -r requirements.txt
  5. Run the app: use $ uvicorn main:app in the bash.
  6. Check the FastAPI: Check the API at the /predict/docs

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.