Tracking using sensor data

摘要

Tracking using sensor data is described, for example, where a plurality of machine learning predictors are used to predict a plurality of complementary, or diverse, parameter values of a process describing how the sensor data arises. In various examples a selector selects which of the predicted values are to be used, for example, to control a computing device. In some examples the tracked parameter values are pose of a moving camera or pose of an object moving in the field of view of a static camera; in some examples the tracked parameter values are of a 3D model of a hand or other articulated or deformable entity. The machine learning predictors have been trained in series, with training examples being reweighted after training an individual predictor, to favor training examples on which the set of predictors already trained performs poorly.

主权项

1. A computer-implemented method comprising:
receiving empirical sensor data from a capture device; applying portions of the sensor data to a plurality of trained machine learning systems to calculate a plurality of diverse predictions of parameters of a process describing how the sensor data is generated; selecting at least one prediction from among the plurality of diverse predictions on the basis of a reconstruction error, where the reconstruction error for each prediction is calculated by generating a signal using the prediction and the process, and calculating a distance between the generated signal and sensor data; and outputting the selected prediction to a system for controlling a computing device using the selected prediction and the process.