Datasets:

---

Sharpa-Robotics

/

SharpaDex-v1.0

Like 59

Follow

Sharpa 21

Hi, thanks for releasing SharpaDex!

We are integrating the dataset and would like to express both observed and commanded end-effector poses in the head-left camera coordinate frame at the current observation timestep.

Could you please clarify the following?

1. Camera calibration
   Are head-camera intrinsics, distortion parameters, and camera-to-head extrinsics available? Do they vary by robot or collection season? Have the released images been rotated, cropped, or rectified?
   A reference FK script or an example showing how to transform a TCP pose into the head-left optical camera frame would be especially helpful.
2. URDF
   What robot platform and arm models were used for collection? Is a matching URDF/MJCF available, including the arms, torso, neck, and camera mounting transforms?
3. TCP pose representation
   For observation.state.tcp_pose and action.tcp_pose, what are the position units and rotation representation? If Euler angles are used, what axis order and intrinsic/extrinsic convention apply?
4. TCP reference frame and origin
   Are the poses expressed relative to a common robot base, each arm’s base, or a camera? Does the TCP origin correspond to the flange, wrist, palm, or another point? How are its local axes defined?

Thank you!