Reflex is introduced as a humanoid catching system that works in roughly one second
A post introducing Reflex says it uses only egocentric RGB-D, without motion capture or teleoperation.
TLDR
A post introducing Reflex claims it enables whole-body humanoid catching in roughly one second using only egocentric RGB-D. The author says it needs neither motion capture nor teleoperation and contrasts the project with academic advice to avoid problems future GPT or Claude models might solve.
Combined views
197
1 Source, first seen ago