A novel application for covid detection using a deep learning web-service.

Create an accurate chest-ct COVID detection service and make it publicly available in a few lines of code.

Source codes for the paper: A Novel and Reliable Deep Learning Web-Based Tool to Detect COVID-19 Infection from Chest CT-Scan The paper Achieved up to 90% accuracy on a dataset of about 750 CT images. Please kindly cite the article at https://arxiv.org/abs/2006.14419, if you find this useful to your application.

NOTE: The Code is Changed to Match the Updated COVID-CT repository, the accuracy might have a +-2 variance.

You are free to change the code to use your desired models.

Full dataset is available on https://github.com/UCSD-AI4H/COVID-CT

Some samples are shown below:

Colab:

Local Jupyter: Notebook Version

git clone https://github.com/KiLJ4EdeN/COVID_WEB
cd COVID_WEB
chmod +x server.sh
./server.sh

The ngrok host url should be displayed.

Note that you can comment out flask ngrok if you dont have an internet connection.

Upload An Image:

Get the Results:

pip3 install bayesian-optimization
python3 run_bayesian_optimization.py
python3 evaluate_model.py

Consider that this is just for demonstration, since the model is already using optimal parameters.

Citation:

@article{Saeedi2020ANA, title={A Novel and Reliable Deep Learning Web-Based Tool to Detect COVID-19 Infection from Chest CT-Scan}, author={Abdolkarim Saeedi and Maryam Saeedi and Arash Maghsoudi}, journal={ArXiv}, year={2020}, volume={abs/2006.14419} }