Datasets:

Project-AgML
/

mango_fruit_stem_detection

dataset_info:
  features:
    - name: image
      dtype: image
    - name: objects
      struct:
        - name: bbox
          list:
            list: float64
        - name: categories
          list:
            class_label:
              names:
                '0': fruit
                '1': stem
    - name: scene
      dtype: string
  splits:
    - name: train
      num_bytes: 7232592817
      num_examples: 1782
  download_size: 7232745055
  dataset_size: 7232592817
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - object-detection
size_categories:
  - 1K<n<10K

Mango Fruit Stem Detection

This dataset provides real-world RGB images of mango fruit and stems captured in agricultural field environments across Hainan Province, China. Images were collected using handheld smartphones during the December 2024 harvest period, reflecting diverse natural conditions for object detection in orchard settings. The dataset contains 1,782 images with 13,785 bounding box annotations across 2 categories.

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

Citation

@article{lu2026feature,
  title={A Feature Enhancement Framework for Joint Mango Fruit and Stem Detection in Complex Orchard Environments},
  author={Lu, Jiahuan and Deng, Qihan and Zheng, Weiping and Cai, Binglong and Zeng, Shan and Li, Jiehao},
  journal={Agriculture},
  volume={16},
  pages={1888},
  year={2026},
  publisher={MDPI}
}

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