Datasets:
Any ETA for the Camera-to-Base Conversion Script?
Thanks for the great work on AgiBotWorld!
In an earlier response, it was mentioned that the URDF file and a script to convert camera extrinsics to the robot base frame (with timestamp alignment) would be released by mid-January. Since it’s already May, may I ask if there’s any update on the release of this script?
Hi! We sincerely apologize, but due to some business considerations, we haven’t been able to upload the URDF at this time.
Regarding the script you mentioned — for converting camera extrinsics to the robot base frame with timestamp alignment — we’ve actually updated the dataset to include frame-by-frame camera extrinsics directly, so that users can work with the data more easily. As a result, we haven’t provided that specific script separately. Thank you for your understanding!
Hi, thanks for the update! I’ve noticed the extrinsics are now aligned per frame — really appreciate AgiBot World’s great open-source work!
Just a suggestion: could the sample_dataset in the alpha release be updated as well? Since the full dataset is quite large, many users (myself included) rely on the sample dataset for testing, but its outdated structure causes some issues.
Really appreciate you pointing that out — it's a great suggestion！We’ll do our best to update it sometime next week when we have a bit of time.
Hi, thanks again for the great dataset and updates!
I noticed that the /state/end/position and /state/end/orientation fields represent the flange pose of the robot. To compute the gripper (effector) position and orientation in world coordinates, may I ask if there is a known fixed transformation from the flange frame to the gripper tip? If so, could you kindly share the translation and rotation (e.g., as a vector or matrix)? That would help a lot with accurate grasp point estimation.
@heealenrens
I agree we should be able to get the gripper position from the extrinsics. However, I suspect the extrinsic data is inaccurate, which prevents a precise calculation. This issue is also mentioned here: https://huggingface.co/datasets/agibot-world/AgiBotWorld-Alpha/discussions/34#684acc2964117289b61f183b
Hoping for an official fix.
@CarbonLeo Thank you very much for your reply. I'm also looking forward to an official fix soon. But I find some detail about their's gripper, Its type is changingtek's CTAG2F120S,and it's z expand length is 220mm-248mm depend on gripper's open range . By the way, have you noticed any issues with the provided camera intrinsic parameters in many cases? I've verified some of them, for example, task_365's episode 652528 is correct, but same task's eposide 657669 is wrong, there are obvious distortion in depth backproject points .