[R] Would you keep a robot demonstration if hand tracking missed the moment the plug went in?

A discussion on robot learning demonstrations questions whether to keep a recording if hand tracking misses a crucial moment, like plugging in a cable. The MEgoVista method is highlighted as a useful starting point, with its detection precision, recall, and F1 scores reported in Table 3. An evaluation protocol in Section 4.4 assigns errors to missed detections. The HaPTIC method failed to produce valid output in multi-person capture scenes, as noted in a comparison against motion-capture ground truth, where all baselines were re-run and rescored on the dataset.

This discussion uniquely highlights the challenge of tracking dropouts in robot learning demonstrations, unlike typical evaluations that exclude such events.