A visual encoder pre-trained on diverse human videos with contrastive and language objectives improves simulated robot manipulation success by over 20% versus training from scratch and enables real Franka arm tasks from 20 demonstrations.
//arxiv.org/abs/2203.13880
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.RO 2representative citing papers
citing papers explorer
-
R3M: A Universal Visual Representation for Robot Manipulation
A visual encoder pre-trained on diverse human videos with contrastive and language objectives improves simulated robot manipulation success by over 20% versus training from scratch and enables real Franka arm tasks from 20 demonstrations.
- From Video to Control: A Survey of Learning Manipulation Interfaces from Temporal Visual Data