Deep learning models have been used to improve seismic trace interpolation in recent years. Encode-Decode models, such as U-Net, have been implemented to solve interpolation problems. The success of U-Net in seismic interpolation has inspired us to test the U-net also as an image generator for further Generative Adversarial Network (GAN) interpolation models. The objective of the present paper is to compare the performance of U-Net inside the GAN models for seismic interpolation: the U-Net alone and two GAN models, the conditional GAN (cGAN) and the Cycle Consistent GAN (CycleGAN), both using the U-Net as a generator inside their workflow. We test the methodologies for two scenarios: regular and irregular interpolation. All tests were performed in a real dataset from the Tupi Field which belongs to the Brazilian pre-salt region. A comparison of the statistical metrics shows that cGAN performs better than CycleGAN and the U-Net alone in most cases. The computational training time of the cGAN model, for all interpolation scenarios, is better than the CycleGAN. Finally, the cGAN training time is comparable to the training of the U-Net alone.
The application of machine learning in geophysics has steeply increased in the last decade, with the quality of its results varying according to the type of seismic problem in focus and the employed computational method. Deep Learning methods are achieving impressive results in this area, but we note that there is still a lack of certainty on whether classical machine learning methods can provide similar results. In the present paper, the objective was to attempt to fill part of that gap, by comparing a well-known non-DL machine-learning method with a DL method for the direct wave propagation and the seismic inversion problems for 2D horizontally-layered models. Both methods are evaluated in different scenarios, but under similar conditions, so that it is possible to understand the effect of parameter configuration on their final results. The dataset has 20,000 samples, each consisting of three vectors: a velocity vector with 236 values (representing a vertical profile of a randomly generated 2D layered model), a reflective vector with 600 values obtained directly from the velocity vector, and the associated seismogram vector with 11 traces, each containing 600 values. The overall results show that the WaveNet produces a lower Mean Squared Error between predicted and correct outputs than that of XGBoost. One challenge yet not dealt with is that the WaveNet can train well in GPU, but we did not succeed in doing the same with the XGBoost, due to the amount of data to be processed.