Improving Single-Image 3D Reconstruction of Porous Media Using Generator with Four-Region Loss Cuts

With the improvement of computer performance, the use of deep learning technology for 3D digital rock core reconstruction has gradually become a hot topic. However, there still remain significant challenges in achieving high-quality imaging using single-slice reconstruction for 3D rock core images. This paper proposes a method that combines four-region random cutting to enhance the performance of single-image reconstruction for 3D images. The following is the structure arrangement of this paper: In the first introduction section, we introduce the background and significance of 3D digital rock cores, and review several representative generative network models. In the second method section, we detail the ease of use of the method adopted in this paper.In the third part, we conduct rigorous experiments to demonstrate the effectiveness of the improved method proposed in this paper. Experimental results show that the method proposed in this paper is effective, greatly improves the training speed and reconstruction quality.