Fig. 4
From: Application of generative adversarial networks (GAN) for ophthalmology image domains: a survey
Examples of problems encountered using GAN techniques. a Mode collapse where the generator produces limited varieties of samples. b Spatial deformity due to small training images without spatial alignment. c Unintended changes due to the difference of data distribution between two domains. d Checker-board artifacts in synthetic images. All of the images were generated according to publicly available datasets and the standard GAN methods