This is a demo for detecting and tracking multiple faces in real time.
It uses the jsfeat CV library for the detection and the tracking.
- We start by running face detection algorithm on the current video frame.
- For each detected face, tracking points are randomly selected within the detected face region.
- We then track these points in the video using an optical flow algorithm, until a specified time interval has passed.
- We then run the detection algorithm again and match the freshly detected faces to the existing ones using a minimum distance approach.
- If a face was not re-detected, we prune it after several detection cycles.
This demo uses getUserMedia to access the camera feed and should work on all browsers except IE.