eloise54/cots_yolo_dataset

CSIRO Crown-of-Thorns Starfish (COTS) Detection Dataset — YOLO Format

This dataset is a modified version of the CSIRO COTS and COTS Scars Dataset, originally released under the Creative Commons Attribution 4.0 License (CC BY 4.0).

Please cite the original dataset authors if you use this dataset in your work.

🙏 Citation

@dataset{csiro_cots_2024,
  author = {Armin, Ali and Bainbridge, Scott and Page, Geoff and Tychsen-Smith, Lachlan and Coleman, Greg and Oorloff, Jeremy and Harvey, De'vereux and Do, Brendan and Marsh, Benjamin and Lawrence, Emma and Kusy, Brano and Hayder, Zeeshan and Bonin, Mary},
  title = {COTS and COTS scar dataset},
  year = {2024},
  publisher = {CSIRO},
  version = {v1},
  doi = {10.25919/03a7-hn83},
  url = {https://data.csiro.au/collection/csiro:64235}
}

🪸 About the Dataset

The original dataset contains images and annotations for Crown-of-Thorns Starfish (COTS) and COTS scars, collected to support coral reef monitoring and control efforts on the Great Barrier Reef (GBR). These starfish are coral predators, and their outbreaks can severely damage reef ecosystems.

Example of use with finetuned yolo11m

⚡ Demo on Hugging Face Spaces:

Gitlab repository containing full training pipeline:

• original CSIRO COTS and COTS Scars Dataset download
• modifications done to make this current dataset
• yolov11m finetuning
• gradio app and inference using this current dataset

🔧 Modifications

Here is the list of the modification made on the original CSIRO COTS and COTS Scars Dataset:

• Annotations converted to YOLO format
• Created train/val split using only images with COTS labels
• Retained the original test set
• Removed all COTS scars annotations
• Organized directory structure as follows:

cots_yolo_dataset/
├── images/
│   ├── train/
│   ├── val/
│   └── test/
├── labels/
│   ├── train/
│   ├── val/
│   └── test/
└── cots_yolo_dataset.yaml

How to use this dataset ?

Training

git clone https://huggingface.co/datasets/eloise54/cots_yolo_dataset
pip install ultralytics

In a python file:

from ultralytics import YOLO

# Dowload pretrained model
import urllib.request
url = 'https://github.com/ultralytics/assets/releases/download/v8.3.0/yolo11m.pt'
urllib.request.urlretrieve(url, 'yolo11m.pt')

# Load a pretrain model
model = YOLO('yolo11m.pt')
results = model.train(data='data/cots_yolo_dataset/cots_yolo_dataset.yaml', epochs=20, task='detect')

Inference without downloading everything

notebook link