The research concludes that edge detection networks can predict more continuous faults than line segments for 2D fault detection; however, additional training data is required.
The paper's code is accessible on GitHub as well as the dataset on google drive.
The paper: "Yu An et al., 2021, Deep convolutional neural network for automatic fault recognition from 3D seismic datasets"
Link for the paper: https://www.sciencedirect.com/science/article/pii/S0098300421000807
Link to the GitHub code: https://github.com/anyuzoey/CNNforFaultInterpretation
Link to the dataset: https://drive.google.com/drive/folders/1JBsDGY4-KTleo0SyYbJ4JaWxzSdeFPU1
Link to the LinkedIn post