Full-Waveform Inversion Method Based on AttInverNet

With the increasing complexity of underground structures in seismic exploration, accurate subsurface velocity modeling has become crucial. Full-Waveform Inversion (FWI) is an effective method for inverting high-precision velocity models, but traditional FWI faces challenges in stability, accuracy, and computational efficiency. With advances in deep learning, this paper proposes a Full-Waveform Inversion method based on experiments on the OpenFWI dataset, which show that AttInverNet outperforms InversionNet in evaluation metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Structural Similarity Index (SSIM), and Single Seismic Inversion Time (SSIT). Once trained, the model can quickly generate subsurface velocity models, avoiding the iterative process and initial model dependency of traditional methods. This makes it particularly suitable for rapid large-scale velocity model construction.