Body
Request to start the recording of an episode.
Passing a dictionnary will store the value in each row of the recorded dataset. The key is the name of the column, and the value is a list. If set to None, no additional metadata is saved.
{ "bbox_position": [0.5, 1, 0, 0.5] }
Path to the branch to push the dataset to, in addition to the main branch. If set to None, only push to the main branch. Defaults to None.
List of camera ids to record. If set to None, records all available cameras.
[0, 1]
Name of the dataset to save the episode in.If None, defaults to the value set in Admin Configuration.
"example_dataset"
Enable rerun
Format to save the episode.
json
is compatible with OpenVLA and stores videos as a series of npy.
lerobot_v2
is compatible with lerobot training..If None, defaults to the value set in Admin Configuration.
json
, lerobot_v2
, lerobot_v2.1
"lerobot_v2.1"
Records steps of the robot at this frequency.If None, defaults to the value set in Admin Configuration.
30
A text describing the recorded task. If set to None, defaults to the value set in Admin Configuration.
"Pick up the orange brick and put it in the black box."
Serial numbers of the leader arms used during the recording
["/dev/ttyUSB0"]
List of robot serial ids to ignore. If set to None, records all available robots.
["/dev/ttyUSB0"]
Record cartesian positions of the robots as well, this will make your dataset incompatible with lerobot and it only works for robots with simulators. Defaults to False.
Target video size for the recording, all videos in the dataset should have the same size. If set to None, defaults to the value set in Admin Configuration.
2
elements[320, 240]
Codec to use for the video saving.If None, defaults to the value set in Admin Configuration.
avc1
, hev1
, mp4v
, hvc1
, avc3
, av01
, vp09
, av1
"avc1"