Table 1 Test set results of the well pad detection pipeline for the Permian and Denver basins individually and together (“Overall”)

From: Deep learning for detecting and characterizing oil and gas well pads in satellite imagery

 

Average precision (mean ± SD)

Precision (mean ± SD)

Recall (mean ± SD)

Permian

0.959 ± 0.002

0.975 ± 0.003

0.906 ± 0.005

Denver

0.928 ± 0.003

0.935 ± 0.005

0.901 ± 0.007

Overall

0.944 ± 0.002

0.955 ± 0.004

0.904 ± 0.006

  1. Because positive examples were downloaded centered around well pads, we used stochastic data augmentations (random cropping and scaling) when evaluating the model to simulate the distribution of well pads during deployment to the full basins, where well pads are rarely centered in the image. As such, the results are reported as the mean and standard deviation (SD) across 10 runs.
Back to article page