Supporting Technologies
Landmark Detection
Faces must be geometrically aligned before they can be compared. We align faces on the basis of anatomically-defined points (e.g. centers of the eye cavities, nose base, etc.) on the face. Our landmark detection technology is able to reliably locate these points across a wide range of face sizes and poses, ranging from frontal to full profile.
more infoHead Pose Estimation
Our face detection technology not only locates faces, but it also estimates the three-dimensional head pose (orientation of the head with respect to the camera). Our head pose estimation technology allows end-users to select between coarse and fine-pose estimation. Coarse-pose detection is more computationally efficient, while fine-pose detection is more accurate in orientation.
more infoVisual Attention
From one frame to the next, videos typically contain a great deal of redundant data. As such, we do not need to analyze every frame in a video sequence in full. Rather, we can focus computational resources on specific “regions of interest” for each frame. Our visual attention technology does this, dramatically speeding up face detection and tracking over videos at virtually no loss in detection and tracking accuracy.
more info