More important, our findings provide
compelling evidence that vision for action works with real-world
metrics, reflecting the actual dimensions of objects, and that
these metrics are computed independently of the scene-based
metrics delivered by vision for perception, which are typically
influenced by contextual cues.