detecting, by a seismic sensor, seismic waves propagated through a subsurface formation;

generating a subsurface image based on the detected seismic waves;

converting, using a geological feature detection neural network, pixels of the subsurface image into probability values, wherein a probability value defines a likelihood of a geological feature occurring at a corresponding location of a pixel in the subsurface formation;

reducing a resolution of the probability values of the subsurface image to create reduced resolution probability values of the subsurface image;

modifying, using distributed weighting factors, one or more of the reduced resolution probability values based on the likelihood of the geological feature occurring at the corresponding location of the pixel in the subsurface formation to generate weighted reduced resolution probability values;

inputting the weighted reduced resolution probability values into a trained generator neural network to generate increased resolution probability values of the subsurface image; and

determining a position of the geological feature in the subsurface formation based on the increased resolution probability values of the subsurface image.