Learnable Gabor kernels in convolutional neural networks for seismic facies classification

Seismic facies classification using a convolutional neural network (CNN) has attracted a lot of attention for its high accuracy. However, its drawback is usually poor generalization when trained with limited training data pairs. Seismic images are dominated by diverse wavelet textures corresponding to seismic facies with various petrophysical parameters, which can be suitably represented by Gabor function. Inspired by this, we propose using learnable Gabor convolutional kernels in the first layer to improve the CNN’s generalization for the task of facies classification. The modified CNN combines the good interpretability of Gabor filters and the reliable learning ability of original CNN. More importantly, it replaces the pixel nature of the CNN filters with a constrained function form that depends on 5 parameters that are more in line with seismic signatures. Further, we constrain the wavelength and angle of the Gabor kernels to certain ranges in the training process based on what we expect in seismic images. The experiments on the Netherland F3 datasets show the effectiveness of the proposed method, especially when applied to testing data with lower signal-to-noise ratios.