A new study from South Korea has proposed a method to determine whether image synthesis systems are producing genuinely novel images, or 'minor' variants on the training data, potentially defeating the objective of such architectures (such as the production of novel and original images). Very oft