Visualize a LeRobot Dataset
SUMMARY
Use LeRobotDataset.visualize() to open a local dataset in Rerun and inspect synchronized camera observations, robot state, and actions.
Method
Load the dataset from its local path before starting visualization:
python
from telekinesis.dataengine import datasets
dataset = datasets.LeRobotDataset(
repo_id="lerobot/aloha_sim_insertion_scripted",
local_path="results/lerobot/aloha_sim_insertion_scripted",
)If the dataset is not available locally yet, follow Load a LeRobot Dataset first.
Parameter Configuration
visualize() has no parameters and returns None.
LOCAL DATASET REQUIRED
The dataset must exist locally; load or download it before visualization.
What to Inspect
Visualization is useful for checking:
- Camera observations across an episode
- Robot state and action changes over time
- Temporal alignment between features
- Unexpected or incomplete demonstrations
- Whether recorded data is suitable for training
Example
python
"""Example script demonstrating how to visualize a LeRobot dataset using the Telekinesis Data Engine."""
from pathlib import Path
from telekinesis.dataengine import datasets
def visualize_lerobot_dataset_example():
"""Load and visualize a LeRobot dataset."""
# 1. Define the dataset identity and local dataset path.
repo_id = "lerobot/aloha_sim_insertion_scripted"
local_path = (
Path(__file__).resolve().parent.parent.parent.parent
/ "results"
/ repo_id
)
# 2. Load the LeRobot dataset from the local path.
dataset = datasets.LeRobotDataset(
repo_id=repo_id,
local_path=local_path,
)
# 3. Visualize the dataset using Rerun.
dataset.visualize()
if __name__ == "__main__":
visualize_lerobot_dataset_example()Next Steps
Inspect Dataset Metadata
Review the feature schema, episode records, statistics, and storage layout.
Inspect metadata →Configure Dataset Loading
Select episodes, temporal windows, transforms, and decoding behavior.
Load dataset →Resume Recording
Append additional demonstration episodes to the existing dataset.
Continue recording →