Figure 1. The overall workflow of multiscale digital rock data fusion based on deep generative adversarial networks: GHR is the generator of StyleGAN2 that takes the style latent vector and random noise as input to simulate SEM images; DHR is the discriminator of StyleGAN2 that aims to distinguish between real SEM samples and fake samples generated by GHR; GL2Hand GH2L are two generators of CycleGAN that predicts high-resolution image from low-resolution micro-CT data and recovers low-resolution data from high-resolution data, respectively; DL and DH are two discriminators corresponding to GL2H and GH2L, respectively.