The Shipwreck-Dataset
The Shipwreck-Dataset is free and proposes images acquired simultaneously by an underwater monocular camera and a multibeam imaging sonar (acoustic camera), both mounted on an ROV (Remotely Operated Vehicle). The images come also with synchronized pressure (depth) and IMU measurements.
The dataset has been acquired on 4 shipwrecks at depths from 50 to 68 meters, in the Mediterranean sea. Thanks to an adjustable mechanical frame, the dataset includes 3 different configurations for the relative distance and orientation of the video camera with respect to the acoustic camera (fixed).
The acquisition has been done in accordance with the UNESCO 2002 Convention on the Protection of the Underwater Cultural Heritage. The experiments have been carried out in 2022, in France, under the supervision of the French Department of Underwater Archaeology (DRASSM – Ministry of Culture).
Acoustic images, monocular vision, IMU and depth
"Basile" ROV
Basile is a 500 m depth rated ROV, developped by LIRMM for the archaeological campaigns of the DRASSM.
During the acquisition of the dataset, it was equipped with Multibeam Imaging Sonar, a monocular video camera, a Razor M0 IMU and a pressure sensor.
video camera
1/3” Megapixel progressive scan CMOS sensor
The camera is installed in an anclosure with a flat window.
acoustic camera
- Frequency 1.2 MHz
- 130° horizontal aperture
- 20° vertical aperture
adjustable frame
This frame is attached under the ROV and allows to control the relative distance and orientation of the video camera with respect to the acoustic camera.
Content of the dataset
The dataset contains 17,572 monocular camera images and 8,577 sonar images. In each folder there is also a sonar_param.txt file containing the range and gain settings of the sonar during the acquisition.
An Excel file summarizes the content of the folders.
Each picture from the camera is associated with the corresponding depth and IMU data, stored in a csv file organized as described below (the camera and sonar indices indicate the indices of the corresponding images’ names) :
depth in m | roll (degree) | pitch (degree) | yaw (degree) | roll speed (m/s) | pitch speed (m/s) | yaw speed (m/s) | sonar index | camera index |
The dataset also contains concatenated pictures, made of an acoustic image on the left and an optical image on the right, both captured synchronously by the system.
How to download the dataset and how to cite us
Nicolas Pecheux, Vincent Creuze, Frédéric Comby et Olivier Tempier, “Self-Calibration of a Sonar–Vision System for Underwater Vehicles: A New Method and a Dataset”, Sensors, 23(3), 1700, Feb. 3, 2023.
* in this paper, you will find more details about the dataset and the description of a self calibration method of the vision-sonar system.
To download the dataset, please click on the button below :
LIRMM : a laboratory of the University of Montpellier (France) and the CNRS (National Center for Scientific Research), Montpellier, France
DRASSM : the National Department of Underwater Archaeology of the French Ministry of Culture
