Huang, Y.B. and Huang, J.P., 2022. Automatic identification of carbonate karst caves using a symmetrical convolutional neural network. Journal of Seismic Exploration, 31: 479-500. Oil and gas reservoirs with cavities are often developed in carbonate rocks. Accurate karst cave identification is an important step in reservoir interpretation. Traditional methods for karst cave detection are generally performed by searching for the beadlike diffraction phenomena in seismic imaging profiles, which are time-consuming and highly dependent on human interactions. We consider the karst cave detection as an image recognition problem of labeling a 2D seismic image with ones on karst caves and zeros elsewhere. We propose an efficient end-to-end convolutional neural network to automatically identify karst caves from the seismic migration images. To train the network, several velocity models are automatically generated first through our self- defined modeling method, and the karst caves are simulated by adding diffraction points. Then these velocity models are transformed into migration imaging results by finite difference method and reverse time migration. The numerical examples show the stability and capability of the proposed network, which is capable of identifying the karst caves even with the seismic data of different qualities and different frequencies. The physical simulated data example also confirms the effectiveness of our method.