Data Export: MCAP to LeRobot

In this lesson, we’ll convert recorded MCAP sessions into a LeRobot dataset for GR00T fine-tuning.

To do that, we’ll:

• Install the MCAP-to-LeRobot converter on the host.

• Convert one or more recording sessions into a single LeRobot dataset.

• Verify that output paths and dataset metadata are valid for training.

Install the Converter

Note

The converter is pure Python and does not require a GPU or a ROS installation. It can run on either an x86_64 workstation or Jetson AGX Thor.

The MCAP-to-LeRobot converter does not require a ROS installation. You can run it either inside the Isaac ROS container or directly on the host.

Tip

If you want to skip data collection and train from precollected data, start from the provided real robot dataset, nvidia/GR00T-N1.7-AppleToPlate, which is already published in LeRobot format.

The asset flow is:

dataset -> training -> checkpoint
dataset + checkpoint -> LEAPP export -> model (ONNX)

hf download \
    nvidia/GR00T-N1.7-AppleToPlate \
    --repo-type dataset \
    --local-dir ${ISAAC_ROS_WS}/recordings/lerobot_output

Install the Converter With uv

The isaac_ros_data_tools repository is cloned during Isaac ROS Setup on Thor.

Note

If you run the converter on an x86 workstation instead of Thor, run the clone step from Isaac ROS Setup on Thor in your workstation workspace before continuing.

1. Install the converter:

   cd ${ISAAC_ROS_WS}/src/isaac_ros_data_tools/isaac_ros_mcap_lerobot_converter
   curl -LsSf https://astral.sh/uv/install.sh | sh
   uv venv --python 3.12 .venv
   source .venv/bin/activate
   uv pip install -e .
   

Convert a Recording Session

1. Run the following script to convert the rosbags of your recording sessions. Replace <session_name> with the name of your recording session.

   To combine multiple recording sessions, pass multiple --rosbags-dir arguments to convert all sessions into a single LeRobot dataset. Episodes are numbered sequentially across sessions in the order listed.

   mcap-to-lerobot \
       --rosbags-dir ${ISAAC_ROS_WS}/recordings/<session_name> \
       --output-dir ${ISAAC_ROS_WS}/recordings/lerobot_output \
       --task "move the apple to the plate" \
       --fps 30 \
       --robot-type unitree_g1
   

   Tip

   mcap-to-lerobot fails if the directory specified by --output-dir already exists.

   In that case, remove the existing output directory first:

   rm -r ${ISAAC_ROS_WS}/recordings/lerobot_output
   

   See also

   Refer to the isaac_ros_mcap_lerobot_converter reference for full CLI details and --fps/sync_rate guidance.

Validate Output for Fine-Tuning

1. Check the dataset directory and metadata before training:

   ls ${ISAAC_ROS_WS}/recordings/lerobot_output
   

2. Confirm the output is ready:

   • Output directory contains four folders: data/, meta/, videos/, and images/.

   • Output directory contains new_embodiment_config_defaults.py for GR00T fine-tuning.

   • data/, meta/, and videos/ contain data.

   • images/ exists but is empty.

   • Episode count aligns with source sessions.

   • fps in metadata matches the recorder sync_rate (30 Hz by default).

Key Takeaways

You converted MCAP sessions into a LeRobot dataset and prepared it for GR00T fine-tuning. The next step is policy fine-tuning on this converted dataset.