The paper says Pi0 is trained with front, wrist, and overhead RGB views. However, the official Pi0 only inputs "base_0_rgb", "left_wrist_0_rgb", "right_wrist_0_rgb". In this way, I'm wondering how you map the three perspectives to these three keys, or do you add new input key (e.g. "front_0_rgb") to contain the front RGB image?
Thanks for you wonderful work again! I'm looking forward to your reply :)
The paper says Pi0 is trained with front, wrist, and overhead RGB views. However, the official Pi0 only inputs "base_0_rgb", "left_wrist_0_rgb", "right_wrist_0_rgb". In this way, I'm wondering how you map the three perspectives to these three keys, or do you add new input key (e.g. "front_0_rgb") to contain the front RGB image?
Thanks for you wonderful work again! I'm looking forward to your reply :)