Posture prediction using recurrent neural network

摘要

PROBLEM TO BE SOLVED: To provide a method of predicting a posture associated with a user in virtual reality or the like. The method receives a request for head posture prediction for an augmented reality experience, identifies a position indicator and a rotation indicator associated with the augmented reality experience, a position indicator and a rotation indicator, and a plurality of cells. Provided to a recurrent neural network (RNN) provided. The RNN can include a plurality of recursive steps, each of which comprises at least one of a plurality of cells and at least one fully connected layer, at least of the plurality of cells. Some are associated with historical periods. RNNs are used to generate posture predictions that correspond to head posture changes for augmented reality experiences for the upcoming period, provide posture predictions in response to requests, and display augmented reality content based on the posture predictions. Trigger. [Selection diagram] FIG. 10