CovidScanBackend: A Python repository from Saransh-cpp

CovidScan Backend

Covid-19 detection using transfer learning.

Description

The API uses the VGG19 convoution neural network, which is trained on a dataset of 21617 images belonging to 2 classes. Number of images used for cross-validation were 6176 and the number of images used for testing were 3089. The classes (as used in the code) -

labels = {
    0: "negative",
    1: "positive",
}

Usage

The API can be accessed through the URL - https://covid-net-backend.herokuapp.com/
To predict an image's class, use the /predict endpoint
For the complete documentation refer to - https://covid-net-backend.herokuapp.com/docs

Running locally

To train the model locally -

Fork and clone the repository

git clone https://github.com/<your_username>/ConvNet-Backend

Create a new virtual environment

python -m venv .venv

Activate the virtual environment

.venv/Scripts/activate

Run the jupyter in the virtual environment

ipython kernel install --user --name=venv
# select the kernel named after your virtual environment in jupyter notebook

To run the API locally-

Fork and clone the repository

git clone https://github.com/<your_username>/SceneNet-Backend

Create a new virtual environment

python -m venv .venv

Activate the virtual environment

.venv/Scripts/activate

Install requirements for training (the Heroku deployment uses tensorflow-cpu and opencv-python-headless because of the memory limitations, but you can switch to tensorflow and opencv-python if you are running this locally)

python -m pip install -r requirements.txt

Fire up the API