YangLiu1021/odom_dataset

Odometry ROSBag Dataset

A multi-sensor ROS1 dataset for odometry research, ToF sensing, inertial analysis, and reproducible Foxglove visualization.

简体中文 · GitHub Repository · Foxglove Sample · Download Guide · Sensor Coverage · Citation

Raw ROSBag files are the primary data source. Lightweight Parquet and YAML indexes provide session-level and topic-level summaries so users can inspect the dataset before downloading large bags. Code, import scripts, validation scripts, tests, and extended documentation are maintained in the GitHub repository: https://github.com/Ly041021/odom_dataset

Highlights

Capability	Included assets
Multi-modal robotic sensing	ToF, flight-controller IMU, LiDAR IMU, MID360 LiDAR, RealSense RGB/depth/infrared camera, odometry, and TF streams
Raw-data preservation	Original ROS1 bag.bag files are retained as the source of truth
Lightweight discovery	Parquet/YAML indexes summarize sessions, topics, duration, coverage, and quality before raw-data download
Reproducible visualization	A Foxglove sample bag and a conversion script are provided for compact inspection workflows
Sensor documentation	Hardware identities, ToF operating mode, calibration status, and quality reports are available in structured files

Quick Facts

Item	Value
Sessions	52
Total duration	12,125.257 sec (202.09 min)
Raw ROSBag size	301.20 GiB
Session index rows	52
Topic index rows	663
Sessions with all indexed sensor groups	26/52
Raw format	ROS1 bag.bag files
Timestamp unit	nanoseconds (ns)
Ready-to-open Foxglove sample	examples/foxglove/visual_demo.bag

Use This Dataset

Download the lightweight indexes first. Avoid cloning the full dataset unless you need all raw ROSBag files.

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="ly041021/odom_dataset",
    repo_type="dataset",
    allow_patterns=["metadata/**", "quality/**", "calibration/**", "dataset.yaml"],
    local_dir="odom_dataset_index",
)

Load the indexes with pandas:

import pandas as pd

sessions = pd.read_parquet("metadata/sessions.parquet")
topics = pd.read_parquet("metadata/topics.parquet")

Download one raw session when you know which bag you need:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="ly041021/odom_dataset",
    repo_type="dataset",
    allow_patterns=["raw/sessions/2026-05-20_030414_odom_run029/**"],
    local_dir="odom_dataset_raw",
)

Download the ready-to-open Foxglove sample:

from huggingface_hub import hf_hub_download

hf_hub_download(
    repo_id="ly041021/odom_dataset",
    repo_type="dataset",
    filename="examples/foxglove/visual_demo.bag",
    local_dir="odom_dataset_visualization",
)

Foxglove Visualization Sample

The dataset includes a ready-to-open Foxglove visualization sample at examples/foxglove/visual_demo.bag. It exposes synchronized RGB, LiDAR, ToF, IMU, and odometry views without requiring users to process a raw bag first.

RGB / Depth	LiDAR

IMU Plot	Odometry Path

ToF Overview

Recommended 3D panel setup: Fixed frame = odom, Display frame = odom.

Data Layout

raw/sessions/<session_id>/bag.bag              # raw ROS1 bag
raw/sessions/<session_id>/bag.bag.sha256       # checksum
raw/sessions/<session_id>/metadata.yaml        # session metadata
raw/sessions/<session_id>/topic_summary.yaml   # per-topic statistics
metadata/sessions.parquet                      # session index
metadata/topics.parquet                        # topic index
metadata/tof_sensor.yaml                       # TOFSense-M cascade metadata
quality/reports/                               # generated quality reports
calibration/robot_v1_template/                 # calibration files
docs/foxglove_visualization.md                 # Foxglove visualization guide
scripts/foxglove_visual.py                     # Foxglove visualization bag builder
examples/foxglove/visual_demo.bag              # ready-to-open Foxglove sample

Sensor Coverage

Sensor group	Session coverage	Notes
LiDAR	44/52	/livox/lidar
RGB camera	28/52	/camera/color/camera_info, /camera/color/image_raw
Depth camera	35/52	/camera/depth/camera_info, /camera/depth/image_rect_raw
Infrared camera	35/52	/camera/infra1/camera_info, /camera/infra1/image_rect_raw, /camera/infra2/camera_info, /camera/infra2/image_rect_raw
ToF	50/52	/nlink_tofsensem_cascade
IMU	52/52	/livox/imu, /mavros/imu/data, /mavros/imu/data_raw
Odometry	52/52	/ekf_quat/ekf_odom, /fusion_odometry/lazy_point_odom
TF	52/52	/tf_static

Observed Topics

Category	Topic	ROS message type	Sessions	Messages	Median Hz
imu	/livox/imu	sensor_msgs/Imu	44	1,977,290	200.00
imu	/mavros/imu/data	sensor_msgs/Imu	52	2,417,806	199.52
imu	/mavros/imu/data_raw	sensor_msgs/Imu	52	2,420,793	199.79
lidar	/livox/lidar	livox_ros_driver2/CustomMsg	44	99,312	10.00
tof	/nlink_tofsensem_cascade	nlink_parser/TofsenseMCascade	50	179,314	14.95
camera	/camera/color/camera_info	sensor_msgs/CameraInfo	28	164,928	29.98
camera	/camera/color/image_raw	sensor_msgs/Image	28	164,925	29.98
camera	/camera/depth/camera_info	sensor_msgs/CameraInfo	35	103,077	15.00
camera	/camera/depth/image_rect_raw	sensor_msgs/Image	35	103,078	15.00
camera	/camera/infra1/camera_info	sensor_msgs/CameraInfo	35	103,073	15.00
camera	/camera/infra1/image_rect_raw	sensor_msgs/Image	35	103,073	15.00
camera	/camera/infra2/camera_info	sensor_msgs/CameraInfo	35	103,072	15.00
camera	/camera/infra2/image_rect_raw	sensor_msgs/Image	35	103,072	15.00
odometry	/ekf_quat/ekf_odom	nav_msgs/Odometry	52	2,395,361	199.31
odometry	/fusion_odometry/lazy_point_odom	nav_msgs/Odometry	51	2,294,922	199.29
tf	/tf_static	tf2_msgs/TFMessage	52	87	1075.35

Quality Notes

• Sessions missing RGB color topics: 24/52.
• Sessions missing ToF topic: 2/52.
• ToF sensor metadata: six-node Nooploop TOFSense-M cascade, UART query mode, 8x8 pixels per node. See metadata/tof_sensor.yaml and metadata/tof_sensor.md.
• ToF validity report: 52,208,469/67,587,392 valid pixels (77.25%).
• quality/reports/topic_completeness_and_tof_validity.md contains the detailed per-session report.
• calibration/robot_v1_template/ contains MAVROS IMU, MID360 LiDAR, D430 infrared stereo camera calibration, and the explicit ToF extrinsic status. Use the recorded /camera/*/camera_info topics for RealSense color/depth intrinsics. Fixed ToF extrinsic is not available in this dataset version.

Build Custom Visualization Bags

scripts/foxglove_visual.py creates Foxglove-ready visualization bags with compressed TOFSense-M overview images, optional RGB topics, standard sensor_msgs/PointCloud2 output at /foxglove/livox/points, accumulated nav_msgs/Path output at /foxglove/odom/path, MAVROS IMU topics, and odometry TF. In Foxglove, use Fixed frame = odom and Display frame = odom for the 3D panel. See docs/foxglove_visualization.md for command-line examples and panel setup.

Limitations

• No fixed train/validation/test split or benchmark protocol is defined in this release.
• Images, depth frames, infrared frames, lidar point clouds, and odometry streams are not exported as separate training files.
• Topic scanning and raw bag decoding require a ROS1 environment with the rosbag Python package.
• Raw bags are large. Start from the Parquet indexes, then download only the sessions you need.

Citation

@dataset{odometry_rosbag_dataset,
  title = {Odometry ROSBag Dataset},
  year = {2026},
  note = {Raw ROSBag storage dataset with metadata indexes}
}

License

Dataset files are licensed under CC BY-NC 4.0. Commercial use is not permitted under this dataset license. See DATA_LICENSE.md for the repository-specific data license note. Source code and documentation tooling in the GitHub repository are licensed separately under MIT.

---

This card is generated from metadata/sessions.parquet, metadata/topics.parquet, and quality/reports/.