Active learning using deep Bayesian networks for surgical workflow analysis.

We show that using DBN-based active learning strategies make it possible to selectively annotate data, thereby reducing the required amount of labeled training in surgical workflow-related tasks.