Walking in place is a standard method for moving through large virtual environments when physical space or positional tracking is limited. This technique has become increasingly prominent with the advent of mobile virtual reality in which external tracking may not be present. In this paper, we revisit walking in place algorithms to address some of their technical challenges. Namely, our solutions attend to improving starting, stopping, and speed control for individual users. From a hand-tuned threshold based algorithm, we provide a new, fast method for individualizing the walking in place algorithm based on biomechanic measures of step rate. In addition, we introduce a new walking in place model based on a convolutional neural network trained to differentiate walking and standing. Over two experiments we assess these methods against a traditional threshold based algorithm on two mobile virtual reality platforms. The assessments are based on controllability, scale, and presence. Our results suggest that an adequately trained convolutional neural network can be an effective way of implementing walking in place.