Utility functions for demo recording and post-processing.
- Folder
record
contains some example bash scripts to record demonstration via rosbag. - Folder
postprocess
contains Python functions and example scripts to extract demonstration data from rosbags and create a HDF5 and a LeRobot dataset.
Build docker image and run the container using the utility scripts build_docker.sh
and run_docker.sh
.
The postprocess
folder will be mounted on the container and it is the container's start directory.
Please make sure to save the demonstration rosbags in a dedicated folder inside postprocess
.
The script manip_demo_rosbag_to_hdf5.py
is an example to convert manipulation demos (inside the folder <demo-name>
) into a HDF5 file.
python manip_demo_rosbag_to_hdf5.py --folder <demo-name>
This will create a HDF5 file named <demo-name>.h5
.
This can be used to create a LeRobot dataset, using the script inria_tiago_hdf5_to_lerobot.py
.
python inria_tiago_hdf5_to_lerobot.py --hdf5_path <demo-name>.h5 --repo_id <org>/<dataset-name> --task <task-name>
By default this script save the dataset locally (inside HF_LEROBOT_HOME=/postprocess/lerobot
).