fig3
From: Unlocking the head: unleashing deep learning and depth camera for free head movement gaze estimation
Figure 3. GPN structure. The 8-dimensional position/posture representation comprises the 3D spatial coordinates of the left and right eyes (6 dimensions) and two angular components of the facial-plane normal representing head pitch and yaw (2 dimensions). GPN: Gaze-Point-Net; FC: fully connected.
