The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
ROS Bag Dataset of Underwater Pool from Narwhal AUV UGM
Dataset Details
Dataset Description
This dataset contains synchronized ROS/ROS 2 bag files recorded during underwater pool tests and navigation tasks by the Narwhal Autonomous Underwater Vehicle (AUV) from Universitas Gadjah Mada (UGM). It includes MAVROS telemetry, state estimation, IMU/pressure sensor readings, internal diagnostics, dynamic frame transformations, and vision feeds targeting underwater gate navigation.
- Curated by: Programmer of Narwhal AUV Team, Universitas Gadjah Mada (UGM): Abyan Raditya and Falah Naufal Zaki and Dzulfikar Saifuddin
- Shared by: Narwhal AUV Team
- Language(s): English (topic metadata and code comments)
- License: MIT
Dataset Sources
- Repository: https://huggingface.co/datasets/narwhal-auv/bawah-air
- Paper: N/A
- Demo: N/A
Uses
Direct Use
- Benchmarking underwater gate detection and perception algorithms.
- Simulating and testing state estimation pipeline performance with MAVROS telemetry.
Out-of-Scope Use
- Open-ocean or high-turbidity navigation without adaptation, as the environmental parameters reflect pool testing conditions.
- Direct deployment on non-MAVROS or non-ROV/AUV architectures without topic remap configurations.
Dataset Structure
The dataset consists of raw .db3 / .bag recording files structured into several primary ROS topic namespaces:
Vision & Target Detection (/gate/*, /narwhal_cam/*)
/narwhal_cam/image_raw(sensor_msgs/msg/CompressedImage): Raw onboard camera video stream./gate/annotated_image(sensor_msgs/Image): Vision pipeline debug stream with bounding boxes/overlays./gate/is_detected(std_msgs/Bool): Detection status flag for the underwater gate./gate/centroid(geometry_msgs/Point): Pixel location of the detected target center./gate/offset(geometry_msgs/Point): Position error relative to image frame center./gate/distance(std_msgs/Float32): Estimated Euclidean distance to the gate./gate/position_3d(geometry_msgs/PointStamped): Estimated 3D coordinate of the gate relative to the vehicle frame.
Vehicle State & Telemetry (/mavros/*)
/mavros/state(mavros_msgs/State): MAVROS connection status, flight mode, and arming state./mavros/imu/data,/mavros/imu/data_raw(sensor_msgs/Imu): Orientation, angular velocity, and linear acceleration./mavros/local_position/odom(nav_msgs/Odometry): Filtered 3D position and linear/angular velocity./mavros/local_position/pose(geometry_msgs/PoseStamped): Local position and orientation./mavros/battery(sensor_msgs/BatteryState): System voltage, current, and battery metrics./mavros/imu/static_pressure,/mavros/imu/diff_pressure: Pressure sensors for depth calculation./mavros/setpoint_raw/*: Desired position, attitude, and trajectory targets.
System & Transforms (/tf, /diagnostics, /uas1/*)
/tf,/tf_static(tf2_msgs/TFMessage): Dynamic and static coordinate transformations between vehicle frames (base_link,camera_link,odom)./diagnostics(diagnostic_msgs/DiagnosticArray): System hardware and node health status./uas1/mavlink_sink,/uas1/mavlink_source: Raw MAVLink packet streams.
Dataset Creation
Curation Rationale
Collected to facilitate research in underwater autonomous navigation, visual perception for competition tasks (e.g., SAUVC), and sensor fusion testing for AUV development.
Source Data
Data Collection and Processing
Data was recorded live in an testing pool environment using standard ROS bag recording utilities. Sensor data streams were synchronized in real-time on the vehicle's onboard companion computer.
Who are the source data producers?
The telemetry and sensor data were generated by the physical sensors onboard the Narwhal AUV (UGM), including the camera system, Pixhawk unit, pressure depth sensors, and companion software nodes.
Annotations
Annotation Process
Gate bounding boxes, centroids, offsets, and 3D position estimates were automatically generated in real-time by the vehicle's onboard vision pipeline nodes (/gate/*).
Who are the annotators?
Automated computer vision detection nodes running onboard during dataset capture.
Personal and Sensitive Information
None. The dataset contains only robotic sensor streams, vehicle telemetry, and underwater pool footage.
Bias, Risks, and Limitations
- Environment Constraints: Data was captured inside a controlled pool environment; lighting conditions, water clarity, and acoustic reflections differ from natural marine environments.
- Hardware Dependencies: Telemetry topics adhere to MAVROS message conventions and ArduPilot message structures.
Recommendations
Users should account for pool wall reflections and static lighting when training visual models on this dataset before deploying to open water.
Citation
@misc{narwhal_auv_ugm_pool_dataset_2026,
author = {Abyan Raditya and Falah Naufal Zaki and Dzulfikar Saifuddin},
title = {ROS Bag Dataset of Underwater Pool from Narwhal AUV UGM},
year = {2026},
publisher = {Hugging Face},
howpublished = {\url{[https://huggingface.co/datasets/narwhal-auv/bawah-air](https://huggingface.co/datasets/narwhal-auv/bawah-air)}}
}
- Downloads last month
- 164