DeepRay is an implementation of medical image Semantic Segmentation using a U-Net architecture. It can detect the infected areas in a COVID-19 patient's chest X-ray image.
I used the Qata-COV19 dataset. The researchers of Qatar University and Tampere University have compiled the QaTa-COV19 dataset that consists of 9258 COVID-19 chest X-ray images.
The U-Net architecture can classify an image down to pixel-level, i.e. Semantic Segmentation. This capability is mainly because of the transposed convolution operator which can up-sample the learned features back to the input's dimension. Here's the architecture:
You may also download the notebook from this repository.
If there are any questions or recommendations, you can reach out to me at itsrezamansouri@gmail.com or std_reza_mansouri@khu.ac.ir.