GeoVolDiff generates synthetic 3D geological volumes with a latent diffusion model trained on physics simulations and shows that networks pre-trained only on this data achieve competitive seismic impedance inversion performance on synthetic and field data.
Ca-diffseg: Cross- attention guided diffusion model for seismic facies segmentation.IEEE Transactions on Geoscience and Remote Sensing, 64:1–15, 2026
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.geo-ph 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GeoVolDiff: Taming 3D Geological Volumes with Latent Diffusion
GeoVolDiff generates synthetic 3D geological volumes with a latent diffusion model trained on physics simulations and shows that networks pre-trained only on this data achieve competitive seismic impedance inversion performance on synthetic and field data.