In my opinion, when deep learning techniques are involved in any approach; it should be useful, practical and robust. The main difficulty here is to train the ANN, and to find the appropriate datasets.
Dear Mitchell Wachtel i don't know.i am working on kaggle pneumothorax segmentation challenge,as a deep learning starter i am using mask rcnn,unet with efficientnet encoder,densenet and ensembling for that kaggle competition,i have no idea about how to transform a code into a fiji/image j plug in ,sorry!