Dataset Card for SeaStereo-Dataset

SeaStereo dataset containing raw RGB and depth images of diverse underwater scenes with a stereo camera in Blender. Scenes include realistic seafloor environments and submerged everyday objects. Dataset will be used to evaluate stereo vision methods in underwater settings.

• Repository: SeaStereo-Generation
• Paper: LinStereo: Linear-Complexity Global Attention for Multi-Scale Iterative Stereo Matching
• Website: LinStereo

Dataset Details

Dataset Description

The dataset has 1176 configurations with 7 camera paths, 4 camera types, 7 water conditions, 1-2 depths and 3 random arranegments of objects. With 40 rendered frames for each configuration, there will be 40,320 raw RGB and depth pairs.

Category	Options
Camera Paths	Approach and retreat 1 Approach and retreat 2 Arc Horizontal pan 1 Horizontal pan 2 Orbit Top-view pan
Camera Types	Low focal length (1.5mm), low interocular (0.04mm) High focal length (2.5mm), high interocular (0.08mm) Low focal length, high interocular High focal length, low interocular
Water Conditions	Clear: Jerlov, Jerlov IA, Jerlov IB, Jerlov II, Jerlov IC Turbid: Jerlov III, Jerlov 3C
Depths	Clear: 5-10m, 15-20m Turbid: 1.7-2.3m
Objects	3 random selections and arrangements of 3-5 objects

Dataset Structure

The dataset uses folder organisation to index the configuration used for each set of 40 images. For example:

Approach_Retreat Camera/
Focal_1.5mm/
Interoc_0.04m/
Jerlov.001/
Depth_Deep1/
Image0001_L.jpg

Inside each Depth_DeepX or Depth_ShallowX folder are the 40 image pairs of .jpg and .exr files.

Citation

BibTeX

@misc{wang2026linstereolinearcomplexityglobalattention,
      title={LinStereo: Linear-Complexity Global Attention for Multi-Scale Iterative Stereo Matching}, 
      author={Yiran Wang and Oliver Turner and Viorela Ila},
      year={2026},
      eprint={2606.25437},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2606.25437}, 
}

APA

Wang, Y., Turner, O., & Ila, V. (2026). LinStereo: Linear-Complexity Global Attention for Multi-Scale Iterative Stereo Matching. arXiv preprint arXiv:2606.25437.

Dataset Card Contact

Oliver Turner Email: otur3695@uni.sydney.edu.au LinkedIn: here

Yiran Wang Email: ywan0794@uni.sydney.edu.au LinkedIn: here