Attention Detection in WebRTC Video Conferencing (WebRTC Tips by WebRTC.ventures)
Covid has changed the paradigm of how the world functions. More and more environments use video conferencing, from schools to business meetings to large conferences. In many of these settings, it is both important and quite useful to be able to gauge participants’ attention to the matter at hand.
In this presentation, Hamza Nasir explores how we can calculate a person’s attention using Head-Pose estimation. He uses TensorFlow’s face landmark detection library to leverage its deep neural network to collect data points on a human face and use that to estimate the head’s pose in a given frame. We can then use the Yaw, Pitch and Roll of the face to calculate the attention metrics for a given participant.
Hamza demos this functionality in a video conferencing solution built with the Vonage Video APIs.