
This work proposes a method to leverage pretrained latent representations of human hand motion to improve the robustness and sample efficiency of behavior cloning for dexterous manipulation tasks on a robotic hand, eliminating the need for costly teleoperation-based data collection.


coremsg

leveraging-pretrained-latent-representations-for-efficient-few-shot-imitation-learning-on-a-dexterous-robotic-hand


Leveraging Pretrained Latent Representations for Efficient Few-Shot Imitation Learning on a Dexterous Robotic Hand
