Once the 2D and 3D images have been standardized, the ensemble model proceeds to predict and classify the 2D and 3D images obtained in blocks 804 and 814, respectively. In various embodiments, the ensemble model analyzes predictions using separate 2D and 3D prediction models. For example, in some embodiments, various 2D and 3D models may have been trained and stored on a computing device (such as those described for FIGS. 1 and 2). In other embodiments, the 2D and 3D models may be trained at blocks 808 and 818 as part of method 800. The 2D and 3D models may be based on, for example, neural network models, such as convolutional neural network, that are trained using training image data sets, e.g., image data sets depicting driver behavior, as described herein. Other models based on different algorithms are also contemplated for the predictive models described herein, for example, a model based on a Random Forest algorithm, that uses a multitude of decision trees and that can output a prediction based on the computation of using the individual trees, such as averaging the tree values.
