Computer vision system tracks instructors' performance
Being a college lecturer – or at least, being a good one – involves more than just speaking at a podium. You also have to engage the students, and a new dual-camera system is designed to assess just how well instructors are doing so.
Known as EduSense, the technology is currently being developed by scientists at Pittsburgh's Carnegie Mellon University. The hardware simply consists of two wall-mounted off-the-shelf video cameras, one of which faces the students and one of which faces the instructor.
Utilizing Carnegie Mellon's previously-developed OpenPose software, the system is able to continuously assess the body positions of all the students at once. This allows it to monitor factors such as where they're looking within the room, how often they raise their hand, and whether or not their posture suggests that they're paying attention.
EduSense also tracks the instructor, noting things like whether they walk around or just stay behind the podium, and how much they move their head and hands.
Frame rates of the video are kept low, in order not to overwhelm the setup with too much data. It should also be noted that the system doesn't digitally identify individuals, plus the video is processed in real time and then erased.
It is hoped that once the technology is developed further, it could incorporate an app that would allow instructors to gauge and adjust their performance in real time, as they're lecturing. Scientists at Spain’s la Universidad Carlos III de Madrid have developed something similar, in the form of augmented reality glasses that let instructors know whether or not students are understanding what they're saying.