Seismic Lithology Interpretation using Attention based Convolutional Neural Networks

Abstract

Seismic interpretation is essential to obtain infor-mation about the geological layers from seismic data. Manual interpretation, however, necessitates additional pre-processing stages and requires more time and effort. In recent years, Deep Learning (DL) has been applied in the geophysical domain to solve various problems such as denoising, inversion, fault estimation, horizon estimation, etc. In this paper, we propose an Attention-based Deep Convolutional Neural Network (ACNN) for seismic lithology prediction. We used Continuous Wavelet Transform (CWT) to obtain the time-frequency spectrum of seismic data which is further used to train the network. The attention module is used to scale the features from the convolutional layers thus prioritizing the prominent features in the data. We validated the results on blind wells and observed that the proposed method had shown improved accuracy when compared to the existing basic CNN.