This repo contains pytorch models for the segmentation of nine different abnomalities based on kid-1 and as well as on kid-2.
The data augmentation techniques are used with the help of an external library called 'Alubmentation' which uses cv2 rather than PIL, for faster augmentation.
For better initialization for training the models, I used an another library called smp segmentation models pytorch.