Fig. 9From: A survey on GAN acceleration using memory compression techniquesTraining architecture for the work [23] showing per-pixel loss and class-lossBack to article page