EgoVerse
GaTech-RL2
Tools for processing, accessing and learning from egocentric human data for robot learning.
What you can explore
Explore data conversion, training and evaluation with the documented episode format.
Before you use it
Camera intrinsics are mandatory in the current format. Source datasets and robot embodiments vary; the toolkit does not make every human recording a hand-control demonstration.
Original contributors
EgoVerse authors and Georgia Tech RL2 contributors
License / access: Code: MIT · check external data, models and hardware terms separately
Keep exploring
EgoMimic
Joint learning from egocentric human video and teleoperated robot data.
DexWild
Portable human demonstrations and robot data combined to train dexterous manipulation policies.
HoloAssist
Human task demonstrations with instructor dialogue and multimodal observations for interactive assistance.
Editorial reference · checked 2026-09-22