Fig. 8From: A survey on GAN acceleration using memory compression techniquesTraining architecture for the work [27] showing per-pixel loss and gan-loss (a.k.a. adversarial loss)Back to article page