A Machine Learning solution for undershoot areas Machine Learning workflows for seismic data interpolation part 3 of 3 The Delft 3D seismic dataset features a bad data zone in t...
A Machine Learning solution for undershoot areas
Machine Learning workflows for seismic data interpolation part 3 of 3
The Delft 3D seismic dataset features a bad data zone in the shallow section right above the historical city center where apparently no seismic data was acquired. The undershoot area affects an area of approx. 35 x 45 traces to a depth of 400 ms. In this example of an OpendTect Machine Learning application we replaced the affected zone with a Unet prediction of the seismic response. A dedicated 2D Unet of 128 x 128 samples was trained on shallow seismic examples with randomly inserted gaps of up to 50 traces. All examples were taken from unaffected areas. The slider compares the original seismic data with the Unet-infilled data.
Next week we will be organizing a free webinar on Machine Learning workflows for seismic data interpolation.
Be sure not to miss out! Thursday 28 January 4 pm Central European time. Sign up here: https://lnkd.in/dHUgGkD
#OpendTect #MachineLearning
