README.md

---
license: cc-by-nc-sa-4.0
---

# ReefNet Dataset

The ReefNet dataset is a comprehensive, curated collection of annotated coral reef images sourced from the CoralNet platform. It is designed for training and evaluating machine learning models for coral classification on a global scale. ReefNet's consolidated label set and rich metadata facilitate the straightforward extraction of custom datasets to support coral classification tasks across different geographic regions.

## Dataset Structure

The dataset is organized as follows:

- **`All_ReefNet_annotations.csv` (2.7 GB)**  
  Contains all annotations available in the main ReefNet dataset, along with the associated metadata.

- **`Metadata_All_ReefNet_annotations.xlsx` (11 KB)**  
  Contains metadata for the `All_ReefNet_annotations.csv` file, describing the meaning of the column names.

- **`ReefNet_RSG-TestDataset_annotations.csv` (2.42 MB)**  
  Contains the annotations of the Red Sea Global (RSG) test dataset, which is used in Experiment 2 of the main paper. This dataset was collected by Red Sea Global (RSG) in the Al-Wajh lagoon in the Red Sea (26.4°N, 36.4°E) in 2022 and was previously stored as a private source within CoralNet.

- **`ReefNet_RSG-TestDataset_images.zip` (1.21 GB)**  
  Contains the images from the Red Sea Global (RSG) test dataset. This dataset is used in Experiment 2 of the main paper as the test dataset and was collected by Red Sea Global (RSG) in the Al-Wajh lagoon in the Red Sea.

- **`Metadata_ReefNet_by_CoralNet_sources.xlsx` (33 KB)**  
  Provides an overview of the CoralNet sources from which the ReefNet dataset was compiled, including attribution to the original creators and metadata such as geographic location and average image resolution. The first sheet explains the content and column names of the other sheets.

- **`ReefNet_labelmapping.xlsx` (79 KB)**  
  Provides an overview of the label set mapping performed to consolidate the ReefNet dataset from the separate CoralNet sources. The first sheet provides metadata explaining the meaning of the content and column names of the other sheets.

- **`ReefNet_RSG-TestDataset_labelmapping.xlsx` (33 KB)**  
  Details the label set mapping performed for Experiment 2 in the main paper, merging labels from the main ReefNet dataset with the RSG-TestDataset. The first sheet explains the meaning of the content and column names of the other sheets.

- **`Overview_Sources_for_Image_Download.csv` (12.8 KB)**  
  CSV file that can be used with the Python script hosted on GitHub to download the imagery belonging to the ReefNet dataset from CoralNet The columns in this file include:
  
  - **Source**: The name of the source from CoralNet.
  - **URL**: The direct link to the source on CoralNet.
  - **ImagesURL**: The base URL structure for accessing the images associated with each source.
  - **ImagesNumber**: The total number of images with confirmed annotations available for that source.
  - **FirstImageNumber**: The starting index or identifier for the first image in the source.

  This file is crucial for users who wish to download images directly from CoralNet using their own methods or with the provided script, ensuring that all images are correctly linked to their respective annotations.

### Downloading Images and Annotations

To download the full set of images from CoralNet and the associated annotations, please use the download script provided in our GitHub repository. This script is designed to handle the specific structure of image storage on CoralNet and will automate the process for you.

You can find the script and detailed instructions in our GitHub repository here:

- **[ReefNet Image Download Script on GitHub](https://github.com/ReefNet-Project/coralnet_crawler)**

Follow the instructions in the repository to download the images and annotations efficiently. The script ensures that all images are correctly organized and matched with the provided annotations.

## Usage

To use this dataset, please ensure that you download all necessary files, including the images from CoralNet. The annotations are provided in CSV format, which can be easily read and processed using various data analysis tools and libraries.

<!-- ## Citation

If you use this dataset in your research, please cite it as follows:

@dataset{reefnet2024,
author = {Yahia Battach, Brian Owain Nieuwenhuis, Xiang Li, et al.},
title = {ReefNet: A Large-scale, Taxonomically Enriched Dataset and Benchmark for Coral Reef Classification},
year = 2024,
publisher = {Hugging Face},
url = {https://huggingface.co/datasets/ReefNet/ReefNet-1.0}
} -->

## Contact

For any questions or issues regarding the dataset, please contact:

- Yahia Battach
- [email protected]

We hope this dataset aids in your research and contributes to the advancement of marine biology and machine learning applications.