Datasets:

Project-AgML
/

potato_disease_classification

dataset_info:
  features:
    - name: image
      dtype: image
    - name: label
      dtype:
        class_label:
          names:
            '0': Black Scurf
            '1': Blackleg
            '2': Blackspot Bruising
            '3': Brown Rot
            '4': Common Scab
            '5': Dry Rot
            '6': Healthy Potatoes
            '7': Miscellaneous
            '8': Pink Rot
            '9': Soft Rot
  splits:
    - name: train
      num_bytes: 92972501
      num_examples: 3905
  download_size: 93064969
  dataset_size: 92972501
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - image-classification
size_categories:
  - 1K<n<10K

Potato Disease Classification

This dataset contains real RGB images of potato tubers captured in field environments, depicting various disease symptoms affecting potato crops. The natural agricultural context of the image collection provides a realistic foundation for developing and evaluating computer vision models focused on potato disease detection. The dataset contains 3,905 images across 10 classes: Black Scurf, Blackleg, Blackspot Bruising, Brown Rot, Common Scab, Dry Rot, Healthy Potatoes, Miscellaneous, Pink Rot, Soft Rot.
Images per class:

Black Scurf: 49
Blackleg: 60
Blackspot Bruising: 770
Brown Rot: 105
Common Scab: 60
Dry Rot: 1,355
Healthy Potatoes: 815
Miscellaneous: 74
Pink Rot: 57
Soft Rot: 560

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

Citation

@article{gkountakos2026potato,
  title={Potato-Free: Eliminating Model Training for Disease Detection and Localization in Potato Tubers},
  author={Gkountakos, Konstantinos and Pasios, Stefanos and Ioannidis, Konstantinos and Demestichas, Konstantinos and Vrochidis, Stefanos and Kompatsiaris, Ioannis},
  journal={Agriculture},
  volume={16},
  pages={1408},
  year={2026},
  publisher={MDPI}
}

The dataset itself can be cited as:

Islam, S., & Afrin, T. (2025). PotatoCare: Deep learning based potato disease dataset [Dataset]. Mendeley Data. https://doi.org/10.17632/7VM7XSKFG4.2

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