Using AI to understand people’s attention and engagement by Teradata

👉 In order to create an estimate of an Audience Attention score, they first analyze each individual to estimate their attention using head pose estimates. They keep track of the median head poses for each individual over the last 30 seconds, then, classify the individual as attentive (green) vs. non-attentive (red) by considering their current deviation from this median head pose.

👉 Lastly, they create an aggregate “attention score” for the audience by combining all individual attentive vs. non-attentive classification. The attention score is a normalized score between 0 and 1, measuring how many individuals are attentive in a given moment, relative to the minimum and maximum of the scene.