Can Generative Adversarial Networks help to overcome the limited data problem in segmentation?

Loss metrics The implementation of the different losses showed a trend that for larger training datasets CE weighting produced better results, whereas smaller training dataset sizes benefitted from dynamic weighting.

also in 53,322 other papers