Fig. 10From: A survey on GAN acceleration using memory compression techniquesTraining architecture for the work [28, 30] showing Intermediate distillation loss, per-pixel loss, and GAN lossBack to article page