Datasets:

Project-AgML
/

subalpine_forest_segmentation

dataset_info:
  features:
    - name: bands
      dtype: binary
    - name: bands_shape
      list: int64
    - name: bands_dtype
      dtype: string
    - name: band_metadata
      dtype: string
    - name: band_order
      list: string
    - name: mask
      dtype: image
  splits:
    - name: train
      num_bytes: 3046469158
      num_examples: 11336
  download_size: 2325883850
  dataset_size: 3046469158
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - image-segmentation
size_categories:
  - 10K<n<100K

Subalpine Forest Segmentation

This dataset provides real RGB imagery of subalpine forest ecosystems captured in the field using a DJI Mavic 2 UAV equipped with a 1-inch CMOS camera. Collected on June 15, 2020, in Wanglang National Nature Reserve, China, the images depict natural forest structures and ground cover under typical high-altitude temperate conditions. It serves as a field-based resource for semantic segmentation tasks in ecological monitoring and forest analysis. The dataset contains 11,336 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{shi2025evaluation,
  title={Evaluation of a CNN model to map vegetation classification in a subalpine coniferous forest using UAV imagery},
  author={Shi, Weibo and Liao, Xiaohan and Wang, Shaoqiang and Ye, Huping and Wang, Dongliang and Yue, Huanyin and Liu, Jianli},
  journal={Ecological Informatics},
  volume={87},
  pages={103111},
  year={2025},
  publisher={Elsevier}
}

Shi, weibo (2025), “CNN model to map vegetation classification in a subalpine coniferous forest using UAV imagery”, Mendeley Data, V2, doi: 10.17632/d9f4m2735b.2

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