Paper Summary
This paper aims to present a novel application of unsupervised machine learning (UML) to semblance-based velocity picking. Earlier methods have used UML to identify semblance maxima within a single panel. The proposed method creates groups of related semblance peaks from different CDP locations, so that users can work interactively with whole groups that span the entire survey, rather than with individual picks. This allows the user to stay in control of the outcome, while delegating much of the tedious labor to the UML algorithm.