Machine Learning workflows for seismic data interpolation - part 2 of 3 In this experiment we investigate how large a gap of missing traces can be filled with a U-Net interpolat...
Machine Learning workflows for seismic data interpolation - part 2 of 3
In this experiment we investigate how large a gap of missing traces can be filled with a U-Net interpolator. We created a training set with 2D seismic images of 128x128 samples. The examples were extracted from one side of the Delft seismic survey. Next, we modified the examples by blanking a group of adjacent traces in each example. The position of the gap and the gap size (5,10,…,35,40 #s) are chosen at random. The slider shows a line from the blind test area. We compare the original data to data with 4 inserted gaps (10,20,30, and 40#s) and the 2D U-Net infill result.
Don't forget to sign up for the webinar on this topic on January the 28th at 4 pm Central European Time https://lnkd.in/dHUgGkD
#OpendTect #MachineLearning
