Application of Convolutional Neural Network in High‐Resolution Seismic Data Analysis for Classification of Holocene Sediments at Southern Vietnam Offshore

The near‐surface 2D reflection seismic method effectively characterizes the shallow distribution of marine sediments by analyzing the acoustic reflectivity wavefield. Furthermore, 2D seismic reflection boundaries are instrumental in representing underlying geological structures. Driven by the need for rapid and accurate interpretation of seismic reflection events, this study utilizes artificial intelligence. We introduce a novel workflow for seismic amplitude picking that uses a convolutional neural network. Real‐world seismic data, acquired from Southern Vietnam, served as comprehensive training, testing, and validation datasets for the development of our convolutional neural network model. The workflow consists of two phases: (i) quick automatic detection of the zones of water volume and marine sediments, and (ii) support for recognition of underground marine sediment types. The application of this network model to measured seismic data from Bac Lieu province, Vietnam, yielded highly accurate predictions in seabed mapping. The seabed deepens in the northwest to southeast trend. Moreover, Holocene layers located upper the older Pleistocene ones, clearly separated with strong seismic signature compatible with the drillholes.