Fig. 4From: Robust head pose estimation based on key frames for human-machine interactionThree views: (a, b) frontal, (b, e) profile and (c, f) side; of the point cloud of all KFs projected to a common frame before (top) and aJer optimization (bottom)Back to article page