The model is implemented according to the architecture present in the paper, as following:
The datased used was downloaded from the ISBI Challenge website, which includes 30 images for the training set as their respective masks. The test set comprises 30 new images.
The images and masks are pre-processed in order to be ready for feeding the implemented model. Since the dataset is of a small amount, data augmentation is used to implement a more accurate and reliable model.
The model is trained for 5 epochs, with 100 steps per epoch.