Training a neural network to track viewer engagement with non-interactive displays

A computerized system comprising:

• one or more data stores;
• a training image component configured to:
• create a plurality of training input images, wherein each training input image represents a path of engagement over time for a viewer during presentation of a display sequence;
• and
• assign each training input image with an engagement level label representing whether the viewer was actively engaged with the display sequence;
• and
• a training component configured to, based on the plurality of training input images and engagement level labels, train a neural network system to predict an engagement level of another viewer during presentation of the display sequence using sensor data captured during presentations of the display sequence.

One or more computer-storage media having a plurality of executable instructions embodied thereon that, when executed by one or more processors, cause the one or more processors to perform a method comprising:

• creating a plurality of training input images, wherein each training input image represents a path of engagement over time for a viewer during presentation of a display sequence;
• assigning each training input image with an engagement level label representing whether the viewer was actively engaged with the display sequence;
• and
• based on the plurality of training input images and engagement level labels, training a neural network system to predict an engagement level of another viewer during presentation of the display sequence using sensor data captured during presentations of the display sequence.