Figure 2
From: Pedestrian orientation dynamics from high-fidelity measurements
(a, c) Pedestrian trajectories (purple) superimposed to depth snapshots (gray). Orientation estimates and local velocities (directions of motion) are reported, respectively, in red and yellow. We estimate shoulder orientation on a snapshot-by-snapshot basis, considering depth “imagelets” centered on a pedestrian. The sub-panel (b) reports an example of such an imagelet with the \(x-y\) coordinate system considered. We employ instantaneous direction of motion \(\theta _v\) extracted from preexisting trajectory data as training labels for a neural network. This yields a reliable estimator for the orientation \(\theta\), accurate even in cases challenging for humans, like in (c). Due to clothing, arms and body posture, presence of bag-packs or errors in depth reconstruction, the overhead pedestrian shape might appear substantially different from an ellipse elongated in the direction of the shoulders.
