Fig. 13From: A survey on GAN acceleration using memory compression techniquesTraining architecture for the work [35] showing both generator and discriminator distillation and using segmentation network to allow content-aware distillationBack to article page