theflash987/Submerged3D

Submerged3D

• Our newly introduced, Submerged3D, is the first dataset captured in deep underwater environments, highlighting real-world challenges such as limited lighting and severe turbidity.

• Submerged3D consists of 4 scenes, each containing 20 RGB 720p images, focusing on shipwrecks captured in the deep sea.

This dataset supports the research described in RUSplatting.

If you think this dataset/project is helpful, please feel free to cite our paper:

@inproceedings{jiang2025rusplatting,
  author    = {Zhuodong Jiang and Haoran Wang and Guoxi Huang and Brett Seymour and Nantheera Anantrasirichai},
  title     = {RUSplatting: Robust 3D Gaussian Splatting for Sparse-View Underwater Scene Reconstruction},
  booktitle = {Proceedings of the British Machine Vision Conference (BMVC)},
  year      = {2025}
}