Dataset Card for Sea-Undistort

Sea-Undistort is a synthetic dataset for through-water image restoration in high-resolution airborne bathymetry. It contains 1,200 scenes with four 512×512 RGB images per scene: (1) ground/no water, (2) undistorted/no waves, (3) no sunglint, (4) distorted (all effects). Each scene comes with structured per-image metadata describing camera, water, sky/illumination, and seafloor parameters. Images were procedurally rendered in Blender to emulate refraction, wave-induced deformations, turbidity/scattering, and sunglint over diverse shallow-water seabeds.

Dataset Details

Dataset Description

Sea-Undistort a synthetic dataset created using the open-source 3D graphics platform Blender. The dataset comprises 1200 image pairs, each consisting of 512×512 pixel RGB renderings of shallow underwater scenes. Every pair includes a “non-distorted” image, representing minimal surface and column distortions, and a corresponding “distorted” version that incorporates realistic optical phenomena such as sun glint, wave-induced deformations, turbidity, and light scattering. These effects are procedurally generated to replicate the diverse challenges encountered in through-water imaging for bathymetry. The scenes are designed with randomized combinations of typical shallow-water seabed types, including rocky outcrops, sandy flats, gravel beds, and seagrass patches, capturing a wide range of textures, reflectance patterns, and radiometric conditions. Refraction is accurately modeled in both the distorted and non-distorted images to maintain geometric consistency with real underwater imaging physics.

In addition, camera settings are uniformly sampled within specific ranges to ensure diverse imaging conditions. Sensor characteristics include a physical width of 36 mm and effective pixel widths of 4000 or 5472 pixels. Focal lengths of 20 mm and 24 mm are simulated with only the central 512x512 pixels rendered. Camera altitude ranges from 30 m to 200 m, resulting in a ground sampling distance (GSD) between 0.014 m and 0.063 m. Average depths range from –0.5 m to –8 m, with a maximum tilt angle of 5°. Sun elevation angles between 25° and 70°, along with varying atmospheric parameters (e.g., air, dust), are used to simulate different illumination conditions. Generated images are accompanied by a .json file containing this metadata per image. 

Sea-Undistort is designed to support supervised training of deep learning models for through-water image enhancement and correction, enabling generalization to real-world conditions where undistorted ground truth is otherwise unobtainable.

• Authors: Kromer, Maximilian (Producer); Agrafiotis, Panagiotis (Supervisor); Demir, Begum (Supervisor)
• DOI: 10.5281/zenodo.15639838
• Version: v1.0.0
• Zenodo publisher date: June 11, 2025
• Available since: August 6, 2025
• Funded by: European Commission MagicBathy - Multimodal multitAsk learninG for MultIsCale BATHYmetric mapping in shallow waters
• License: CC BY-NC-SA 4.0

Dataset Sources

• Project Page: Sea-Undistort
• Record & Download: Zenodo
• Repository: Github
• Paper: arXiv.org and IEEE

Dataset Structure

Images per scene:

• render_####_ground.png → no_water (seafloor only)
• render_####_no_waves.png → undistorted (minimal surface distortions)
• render_####_no_sunglint.png → no_sunglint
• render_####.png → distorted (all effects)

Metadata file: scene_settings.json contains camera, seafloor, sky/illumination, and water-shader parameters for each scene. In this repository we provide a metadata.jsonl that flattens a useful subset of those fields for 🤗 Datasets.

Dataset Creation

Motivation Real paired images with and without surface/column effects are practically unavailable; Sea-Undistort enables supervised training with physically consistent synthetic pairs.

Generation Blender-based procedural scenes vary seabed materials/geometries, water-shader parameters, solar/atmospheric conditions, and camera settings. Refraction is modeled in both distorted and non-distorted variants; per-image metadata is provided.

Pretrained baselines The authors provide instructions and links for NDR-Restore, ResShift, and ResShift+EF pretrained on Sea-Undistort. These models can be found on Github.

Citation

If you find this dataset useful, please consider giving a like ❤️.
If you use the code in this repository or the dataset please cite:

Kromer, M., Agrafiotis, P., & Demir, B. (2025). Sea-Undistort: A dataset for through-water image restoration in high resolution airborne bathymetric mapping. arXiv. https://arxiv.org/abs/2508.07760

BibTeX:

@misc{kromer2025seaundistortdatasetthroughwaterimage,
      title={Sea-Undistort: A Dataset for Through-Water Image Restoration in High Resolution Airborne Bathymetric Mapping}, 
      author={Maximilian Kromer and Panagiotis Agrafiotis and Begüm Demir},
      year={2025},
      eprint={2508.07760},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2508.07760}, 
}

DOI (Zenodo): 10.5281/zenodo.15639838. License: CC BY-NC-SA 4.0.

Dataset Card Authors

Kromer, Maximilian

Dataset Card Contact

Kromer, Maximilian: m.kromer@tu-berlin.de