- This project is designed for the Sigmahacks 3 hackathon.
- You need to have Python 3.7 or higher and PyQt5.
First, a Jupyter Notebook is created with the algorithm to train the pre-trained model with ImageNet DenseNet-201. You can view it in the model/train_model folder. It was trained with images from an external public access repository.
The interface is created in Qt Creator to achieve a more comfortable interaction for the user, in addition to maintaining the KISS principle.
The model is saved in the model folder and is used in the app.py script, in which the whole PyQt5 structure is implemented.
Finally, it is tested with two internet images, one of the chest with COVID-19 and the other of the normal chest, both of which are X-ray images, in addition to being images never seen before by our model, which successfully predicted both images.
It is a powerful 201-layer convolutional neural network that obtains a more minimal error compared to the human eye. This network was trained with ImageNet, so only in this project we did transfer learning for our two classes [Normal, Covid19].