This dataset is available for non-commercial research use only, under a signed Data Use Agreement. The agreement can be downloaded at agreement link, and we will also e-mail it to you after you submit this form. Access is approved only after the signed copy is returned. By submitting, you confirm that you will not redistribute the data, will not use it for commercial purposes or weapons targeting, and will cite the associated publication.

S3Simulator Synthetic Sonar Dataset (Ship and Plane)

Summary

A dataset of 7,721 synthetic side-scan sonar images of two target classes, ship and plane, generated with S3Simulator and described in the paper cited below. It is intended as a benchmark for underwater image analysis when labeled real sonar data is scarce.

The mine class is released separately in a companion dataset generated with S3Simulator+.

Dataset contents

Class	Images	Nadir-zone variant
ship	3,766	3,766
plane	3,955	3,955
Total	7,721	7,721

Resolution: 1618 x 840 pixels for every image
Format: JPEG
Channels: stored as 3-channel RGB, but the content is single-channel intensity (the three channels are identical apart from JPEG rounding, with a maximum difference of 2 levels out of 255). Load as one channel, for example Image.open(path).convert("L").
Labels: image-level class, given by the folder name

Repository structure

ship/                 ship_00001.jpg ... ship_03766.jpg
plane/                plane_00001.jpg ... plane_03955.jpg
ship_nadir_zone/       ship_nadir_00001.jpg ... ship_nadir_03766.jpg
plane_nadir_zone/     plane_nadir_00001.jpg ... plane_nadir_03955.jpg
splits/                train.txt, val.txt, test.txt
metadata/               filename_map.csv
checksums.sha256

metadata/filename_map.csv maps each released file name to its original file name.
checksums.sha256 holds a SHA-256 hash for every image.

Data splits

Fixed splits are provided in splits/, one image path per line. They were made by a random 80/10/10 split within each class (seed 42), applied to the original images. Each original image's nadir-zone counterpart (ship_nadir_zone/ or plane_nadir_zone/) is always placed in the same split as its original, so the split files list both versions of every image.

Split	Original images	Nadir-zone images	Total entries
train	6,176	6,176	12,352
val	772	772	1,544
test	773	773	1,546
Total	7,721	7,721	15,442

If you only want the original renders (excluding the nadir-zone variant), filter splits/*.txt to paths under ship/ and plane/ only.

How the data was generated

The images were generated with S3Simulator, a side-scan sonar simulator described in the paper cited below.

Nadir-zone variant

ship_nadir_zone/ and plane_nadir_zone/ contain a processed version of every image in ship/ and plane/, with an added simulated sonar nadir-zone effect: Gaussian noise, a bright bordered dark stripe with spike artifacts, and linear brightness gradients applied to approximate the nadir (directly below the sensor) region seen in real side-scan sonar imagery. This is a post-processing augmentation applied to the rendered images, not a separate simulator output.

Each nadir-zone image corresponds 1:1 to an original image by index, e.g. ship_nadir_00001.jpg is the nadir-zone version of ship_00001.jpg. Both versions of a given image are always in the same split (train/val/test), so they do not appear separately across splits.

Loading the data

from PIL import Image
from pathlib import Path

root = Path("path/to/dataset")
train = (root / "splits" / "train.txt").read_text().split()

img = Image.open(root / train[0]).convert("L")   # single channel
label = train[0].split("/")[0]                   # "ship" or "plane"

Verifying the files

Linux and macOS:

sha256sum -c checksums.sha256

Windows PowerShell:

Get-FileHash ship\ship_00001.jpg -Algorithm SHA256

Compare the result with the matching line in checksums.sha256.

Intended use

Sonar target classification and detection research
Sim-to-real transfer and domain adaptation
Data augmentation and low-data learning for underwater imagery
Benchmarking generative models for sonar imagery

Out-of-scope use

Weapons targeting or offensive military systems
Drawing conclusions about real-world operational performance from evaluation on this synthetic data alone

Limitations

Sim-to-real gap. Speckle, acoustic shadows, multipath and seabed clutter in simulated images differ from real sonar. Validate on real data before making any claim about deployment.
JPEG compression is lossy and can alter speckle statistics.
Only two classes are included here (ship and plane), and class counts are not identical.
Labels are image-level. No bounding boxes or masks are provided in this release.

Access and terms

Access is granted on request after manual review. The dataset is released under CC BY-NC 4.0 for non-commercial research use. Please do not redistribute the files; point others to this page instead.

Citation

If you use this dataset, please cite:

@inproceedings{kamal2024s3simulator,
  title={S3simulator: A benchmarking side scan sonar simulator dataset for underwater image analysis},
  author={Kamal Basha, S and Nambiar, Athira},
  booktitle={International Conference on Pattern Recognition},
  pages={219--235},
  year={2024},
  organization={Springer}
}

Authors

Kamal Basha S and Athira Nambiar, SRM Institute of Science and Technology.

Contact

athiram@srmist.edu.in,bashakamal95@gmail.com. Use this address to report misuse or to request removal of the data.

Downloads last month: 28

Total file size:

7.92 GB