Generating Heterogeneous 3D Geological Microstructures from 2D Images via a Stable Diffusion-Adversarial Model
A hybrid diffusion-adversarial model generates heterogeneous 3D geological microstructures from cheap 2D images, replacing denoising loss with adversarial loss for stable training.
The work reconstructs 3D volumes of clay and cementitious materials from 2D images using a denoising diffusion model trained with an adversarial loss, since no 3D ground truth exists. Building on the limitations of GAN-based SliceGAN for complex heterogeneous microstructures, the approach produces volumes with minimal slice artefacts and close agreement with ground-truth phase fractions and structural descriptors. Applications include materials science and geological waste disposal characterization.
- Replaces standard denoising loss with adversarial loss, enabling stable diffusion training without 3D ground truth.
- Generalizes SliceGAN-style 3D generation to heterogeneous microstructures with minimal slice artefacts.
- Generated volumes match ground-truth phase fractions and structural descriptors for clay and cementitious materials.
Full article164 words · extracted from arxiv.org · click to collapse
Characterizing the physical properties of clay and cementitious materials matters across many fields, from materials science to geological waste disposal. Property simulation typically calls for 3D imaging, which is expensive, not always accessible, and technically limited for certain materials. Recent progress in deep generative models offers a way around this, reconstructing 3D volumes from the more easily acquired 2D images. Among GAN-based methods for 3D microstructure generation, SliceGAN has shown strong results for homogeneous isotropic and anisotropic systems. It struggles, however, to capture the finer detail of more complex heterogeneous microstructures, which motivates alternative generative frameworks. We introduce a hybrid approach that draws on the stability and generation quality of denoising diffusion models. Since no 3D ground truth is available, we replace the standard denoising loss with an adversarial loss, which yields a stable training process in our experiments. We show that the resulting model generates microstructures of varying complexity with minimal slice artefacts and close agreement with ground-truth phase fractions and structural descriptors.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.20358