Data Loggers
SUMMARY
Data Loggers capture information produced during robot execution and organize it into reusable datasets and logs for inspection, replay, analysis, and training.
When to Use Data Loggers?
Once robots begin operating, they become a source of new data. Cameras, sensors, controllers, perception systems, and policies continuously produce information that can help reproduce failures, evaluate behavior, and improve future Skills.
Data Loggers preserve runtime data in a format suited to its source and intended use instead of leaving it in disconnected application logs.
How to use Data Loggers?
A logging workflow generally follows four stages:
- Select a logger for the data being produced.
- Capture data while the robot or application runs.
- Finalize the recording in its supported dataset or log format.
- Inspect, replay, analyze, or use the result for training.
The recorded output can become part of a growing training corpus, allowing deployment data to improve the next generation of Skills.
LOGGER OR DATASET ANALYSIS?
Use a Data Logger to capture data during execution. Use Dataset Operations to create, load, inspect, modify, or visualize an existing supported dataset.
Guides
Select the logger that matches the structure and intended use of your data.
Logger
Detection Dataset Logger
Write labeled frames to YOLO or RF-DETR/COCO datasets in real time — directly from a running detection pipeline.
Explore →Logger
LeRobot Dataset Logger
Record episodic robot demonstrations containing synchronized observations, actions, and tasks.
Explore →Logger
MCAP Logger
Record every Zenoh topic on the BabyROS network to a single MCAP file. Replay and read back with one generator call.
Explore →