Datasetpublic

lan374/NPU-SLAM

NPU-SLAM Dataset A LiDAR-Camera-Inertial SLAM dataset built for FAST-LIVO2, with STM32 hardware-triggered synchronization across a Livox Mid360 LiDAR and three cameras. Sensors Livox Mid360 (LiDAR) · built-in IMU · 3× UVC camera (cam0 main, cam1/cam2 auxiliary) Format ROS1 bag rosbag2.0 Frames LiDAR 10 Hz, IMU 200 Hz, Cameras ~10 Hz (640×480) Synchronization STM32 hardware trigger, shared clock License CC BY 4.0 Get the data git clone… See the full description on the dataset page: https://huggingface.co/datasets/lan374/NPU-SLAM.

sourceHugging Faceupdated 23d agoView on Hugging Face

0likes254downloads

Dataset Card

NPU-SLAM Dataset

A LiDAR-Camera-Inertial SLAM dataset built for FAST-LIVO2, with STM32 hardware-triggered synchronization across a Livox Mid360 LiDAR and three cameras.


Sensors	Livox Mid360 (LiDAR) · built-in IMU · 3× UVC camera (cam0 main, cam1/cam2 auxiliary)
Format	ROS1 bag `rosbag2.0`
Frames	LiDAR 10 Hz, IMU 200 Hz, Cameras ~10 Hz (640×480)
Synchronization	STM32 hardware trigger, shared clock
License	CC BY 4.0
Get the data	`git clone https://huggingface.co/datasets/lan374/NPU-SLAM`

All camera frames are published on image_synced with timestamps aligned to the LiDAR frames (hardware-triggered).

Devices

Two capture platforms:

Device	Sequences
<img src="website/data/device_1.jpg" width="220">	seq01 – seq05
<img src="website/data/device_2.jpg" width="220">	seq06 – seq07

Sequences

Covers indoor (scan / corridor) and outdoor (open area / between buildings) scenes for LiDAR-Camera-Inertial SLAM. Preview frames are sample cam0 images.

Preview	Sequence (public / folder)	Scene	Duration	Size
<img src="website/data/seq01indoor1.jpg" width="180">	seq01indoor1 / `seq_01_d1_indoor1`	indoor1	~104 s	3.3 GB
<img src="website/data/seq02indoor2.jpg" width="180">	seq02indoor2 / `seq_02_d1_indoor2`	indoor2	~96 s	3.0 GB
<img src="website/data/seq03indoor3.jpg" width="180">	seq03indoor3 / `seq_03_d1_indoor3`	indoor3	~574 s	18.1 GB
<img src="website/data/seq04outdoor1.jpg" width="180">	seq04outdoor1 / `seq_04_d1_outdoor1`	outdoor1	~201 s	6.3 GB
<img src="website/data/seq05outdoor2.jpg" width="180">	seq05outdoor2 / `seq_05_d1_outdoor2`	outdoor2	~178 s	5.6 GB
<img src="website/data/seq06outdoor1.jpg" width="180">	seq06outdoor1 / `seq_06_d2_outdoor1`	outdoor1	~101 s	2.4 GB
<img src="website/data/seq07outdoor2.jpg" width="180">	seq07outdoor2 / `seq_07_d2_outdoor2`	outdoor2	~171 s	5.4 GB

Each sequence folder contains:

—data.bag — /livox/lidar (CustomMsg), /livox/imu, /camera/cam{0,1,2}/image_synced
—info.yaml — topic counts, rate, duration, scene metadata
—sync_timestamps.csv — sync log
—preview.jpg — a sample cam0 frame

Sensor Configuration

Sensor	Model	Topic	Rate	Resolution
LiDAR	Livox Mid360	`/livox/lidar`	10 Hz	—
IMU	Mid360 built-in	`/livox/imu`	200 Hz	—
Camera cam0	UVC	`/camera/cam0/image_synced`	~10 Hz	640×480
Camera cam1	UVC	`/camera/cam1/image_synced`	~10 Hz	640×480
Camera cam2	UVC	`/camera/cam2/image_synced`	~10 Hz	640×480

Calibration

—calibration/camera_intrinsics.yaml — camera intrinsics
—calibration/lidar_camera_extrinsic.yaml — LiDAR–camera extrinsic (cam0 calibrated; cam1/cam2 pending FAST-Calib)
—calibration/imu_params.yaml — IMU parameters

Citation

If you use this dataset, please cite:

bibtex

@misc{lan374NPUSLAM,
  author = {Lan, ...},
  title = {NPU-SLAM: LiDAR-Camera-Inertial SLAM Dataset},
  year = {2026},
  howpublished = {\url{https://huggingface.co/datasets/lan374/NPU-SLAM}},
}

Usage

The bags are directly in the repository. Play with FAST-LIVO2 (topics match mid360.yaml):

bash

git clone https://huggingface.co/datasets/lan374/NPU-SLAM
cd NPU-SLAM
rosbag play sequences/seq_01_d1_indoor1/data.bag
roslaunch fast_livo mapping_mid360.launch