Qiao, Z.X., Chuai, X.Y., Xu, Z.W., Guo, N.C., Zhu, W., Zhang, J.F., Chen, W. and Xia, R., 2023. Seismic data enhancement based on Bayesian convolutional neural network. Journal of Seismic Exploration, 32: 407-425. The acquisition of high-quality seismic data is an important goal of seismic data processing. Traditional seismic data processing methods are usually used alone to remove noise or improve resolution. They can only improve the quality of seismic data from a certain point of view, and lack the protection of effective detail signals. In order to improve the quality of seismic data from whole angle, protect and highlight the details of geological structures such as faults and fault uplifts, this paper proposes to apply Bayesian Convolutional Neural Network (BCNN) to seismic data processing to enhance seismic data. BCNN is an organic combination of Bayesian theory and neural network, which can avoid network over-fitting and enable the network to learn deeper data features adaptively, with better robustness. In addition, the up-sampling operation at the end of the network model is conducive to preserving the feature information of seismic data in the low-resolution space. In this paper, the seismic data is enhanced based on F3 dataset, and compared with the general full convolutional neural network (FCNN) and construction-oriented filtering methods. The results show that the proposed method can better highlight the structural details, improve the interpretability of seismic data, and is an effective means to enhance fault and uplift structures.