Dataset Viewer (First 5GB)
Auto-converted to Parquet
Search is not available for this dataset
The dataset viewer is not available for this split.
Rows from parquet row groups are too big to be read: 404.75 MiB (max=286.10 MiB)
Error code:   TooBigContentError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Coral reefs 3D photogrammetry

Description

We 3D mapped multiple coral reefs in Indonesia (following this protocol) and sharing all our data with you πŸ€—

exampel image

This dataset currently contains 90,289 (352GB) of raw GoPro images and some colour-corrected images.

Additional data - including camera poses, reconstructed 3D point clouds, 3D polygonal meshes, orthomosaics, annotations, and 3D Gaussian Splatting models - will be added soon. We just decided share raw data right now, and processed as soon we have it, so we don't block the progress.

There's one survey site (farm) has a lot of photos with a ton of coral reef nurseries. It was collected not for reconstruction, but rather for training species classification models, because each table has the same coral species.

Everything (except the farm) is correctly georeferenced and scaled.

We hope this dataset helps improve AI models for coral reef species classification, 3D semantic segmentation, 3D reconstruction, disease detection and help protect these precious ecosystems! πŸͺΈ

Contact: [email protected] or [email protected] if you have any questions.

License

This dataset is licensed under CC-BY-4.0 - no strings attached, even for commercial use, as long as attribution provided. Have fun!

Structure

Each dataset directory includes:

...
indonesia_tabuhan_p2_20250210/   # {country}_{site}_{patch}_{date}/
indonesia_pemuteran_p1_20250213/
...
β”œβ”€β”€ raw/                         # Raw, unprocessed images
β”œβ”€β”€ corrected/                   # Colour-corrected and unwarped images (Pinhole format)
β”œβ”€β”€ colmap/                      # 3D reconstruction in COLMAP format
β”‚   β”œβ”€β”€ cameras.bin              # Camera parameters
β”‚   β”œβ”€β”€ images.bin               # Image data and pose information
β”‚   └── points3D.bin             # 3D point cloud data
β”œβ”€β”€ 3dgs/                        # 3D Gaussian Splatting models
└── metadata.json                # Metadata including GPS coordinates, equipment

File stats

You can see file stats below, along with different categories.

Legend:

  • βœ… Uploaded and available
  • 🟑 Complete but needs uploading
  • βšͺ Not done yet
Survey patch Raw Count Size Corrected Poses 3D Point cloud Ortho 3DGS
Banyuwangi Farm 20250211 βœ… 8,700 33.98 GB βšͺ 🟑 🟑 βœ… βšͺ
Pemuteran P1 20250213 βœ… 10,246 40.02 GB βšͺ 🟑 🟑 βœ… βšͺ
Pemuteran P2 20250213 βœ… 6,980 27.26 GB βšͺ 🟑 🟑 βœ… βšͺ
Pemuteran P3 20250213 βœ… 7,580 29.60 GB βšͺ 🟑 🟑 βœ… βšͺ
Tabuhan P1 20250210 βœ… 6,590 25.74 GB βœ… βœ… βœ… βœ… βœ…
Tabuhan P2 20250210 βœ… 8,822 34.46 GB βšͺ 🟑 🟑 βœ… βšͺ
Tabuhan P3 20250210 βœ… 6,952 27.15 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P1 20250208 βœ… 8,395 31.85 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P2 20250208 βœ… 3,871 15.12 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P3 20250208 βœ… 2,938 11.47 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P4 20250209 βœ… 8,846 34.55 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P5 20250209 βœ… 4,412 17.23 GB βšͺ 🟑 🟑 βœ… βšͺ
Watudodol P6 20250209 βœ… 5,957 23.26 GB βšͺ 🟑 🟑 βœ… βšͺ
Total 90,289 351.75 GB

You can see more info in the metadata.json.

Contributors

This dataset was made possible by the amazing efforts of many brilliant people:

with the massive support of these organisations:

Contact [email protected] or [email protected] if you have any questions about the dataset.

Cite

@misc{wildflow_sweet_corals_indo_2025,
    author       = { Sergei Nozdrenkov and Mujiyanto and Raymon Rahmanov Zedta and Fakhrurrozi and Triona Barker and Muhammad Dziki Rahman and Andrias Steward Samusamu and Fathur Rochman and Leigh Peters and Dr Rita Rachmawati and Dr Ofri Johan and Dr Jamie Craggs and Prof Michael Sweet },
    title        = { sweet-corals (Revision bd316dc) },
    year         = 2025,
    url          = { https://huggingface.co/datasets/wildflow/sweet-corals },
    doi          = { 10.57967/hf/5162 },
    publisher    = { Hugging Face }
}

Links:

Downloads last month
3,451