bagx

Post-processing analysis engine for ROS2 rosbag data.

Try bagx in your browser — drop a .db3 bag, get a readiness report. No install, and your bag never leaves your machine.

bagx is not a replacement for ros2 bag. ros2 bag handles recording & playback (I/O). bagx handles evaluation, understanding & comparison (Analysis).

What can you do with bagx?

# Score a bag's sensor quality
bagx eval recording.db3
bagx rules list
bagx eval recording.db3 --rules warehouse_bot

# Compare two runs
bagx compare run_A.db3 run_B.db3

# Check time sync between sensors
bagx sync recording.db3 /camera /lidar

# Find sensor anomalies
bagx anomaly recording.db3

# Export for ML pipelines
bagx export recording.db3 --ai --format parquet

# Batch-evaluate a dataset
bagx batch eval ./recordings/*.db3 --csv summary.csv

# Re-run a benchmark suite on curated public/private bags
bagx benchmark benchmarks/open_data_suite.json
bagx benchmark benchmarks/non_slam_suite.json

Real-world results

Evaluated on public SLAM datasets (see the scoreboard for all 30 rows and reproduce commands):

Dataset	Duration	Messages	GNSS	IMU	Sync	Overall
Livox MID-360 (AIST driving)	277s	58K	N/A	97.3	100.0	98.7
TUM VI calib-imu1	52s	18K	N/A	90.8	100.0	95.4
NTU VIRAL eee_03 (drone)	181s	390K	80.0	72.0	100.0	84.0
EuRoC MH_01_easy	188s	47K	N/A	23.1	97.6	60.3

Quick install

pip install bagx

See Getting Started for details.