Datasets:

Project-AgML
/

pine_wilt_segmentation

dataset_info:
  features:
    - name: image
      dtype: image
    - name: mask
      dtype: image
  splits:
    - name: train
      num_bytes: 1903512865
      num_examples: 2238
  download_size: 1907953672
  dataset_size: 1903512865
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - image-segmentation
size_categories:
  - 1K<n<10K

Pine Wilt Segmentation

This dataset provides high-resolution aerial RGB imagery of pine forests in Jinju, South Korea, captured in January 2023 for semantic segmentation tasks focused on pine wilt disease detection. Acquired at 5-cm resolution using Korea Aerospace Research Institute (KARI) aerial platforms, the images depict field conditions with varying disease severity across pine tree canopies. The real-world, high-detail aerial data offers a practical resource for developing computer vision models in agricultural disease monitoring. The dataset contains 2,238 images with pixel-level mask annotations.

This dataset is indexed on https://project-agml.github.io/ as part of the AgML python library.

Citation

@article{ha2025enhanced,
  title={Enhanced pine wilt disease outbreak prediction: Integrating deep learning- detected infected trees with species distribution modeling},
  author={Ha, Uirin and Kim, Hyungho and Kim, Seunguk and Choe, Hyeyeong},
  journal={Ecological Informatics},
  volume={91},
  pages={103421},
  year={2025},
  publisher={Elsevier}
}

Ha, Uirin; Kim, Hyungho; Kim, Seunguk; Choe, Hyeyeong (2025), “Enhanced pine wilt disease outbreak prediction: integrating deep learning-detected infected trees with species distribution modeling”, Mendeley Data, V3, doi: 10.17632/swpp8jxymv.3

This dataset was reformatted from its original format to match HuggingFace standards.