SeisReconNO
Neural operators for 3D seismic reconstruction
About
SeisReconNO provides reproducible material for SeisReconNO: Leveraging a U-Net-Enhanced Fourier Neural Operator for 3D Seismic Reconstruction (Traversa et al. 2026). The project addresses missing traces in 3D seismic data using a mesh-invariant reconstruction approach that combines spectral and spatial representations.
Why it matters
Missing seismic traces degrade downstream imaging and interpretation. A research project page should explain that practical problem before directing technical readers to the implementation and paper.
Key result and audience
The project trains on open 3D field cubes with random, trace-wise, and regular missing patterns. The paper reports reconstruction across missing ratios and generalization to unseen data resolutions, making the repository relevant to seismic processors and neural-operator researchers.
Public links
This is an original SEEM-KFUPM repository and is distributed under the MIT License.