---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype:
      class_label:
        names:
          '0': als
          '1': bean_rust
          '2': healthy
          '3': unknown
  - name: split
    dtype: string
  splits:
  - name: train
    num_bytes: 107273198
    num_examples: 8623
  download_size: 107318556
  dataset_size: 107273198
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc0-1.0
task_categories:
- image-classification
size_categories:
- 1K<n<10K
---
# Bean Disease Classification Uganda

The Makerere University bean image dataset used to train BeanWatchNet: smartphone photos of bean leaves collected in farmers' fields in twelve districts across three regions of Uganda, classified as angular leaf spot (ALS), bean rust or healthy, plus an 'unknown' class of unrelated images (other crops, cars, sky, buildings) added for out-of-distribution handling. Experts annotated the images, and the record provides them in training, validation and test splits. The dataset contains 8,623 images across 4 classes: als, bean_rust, healthy, unknown.  
Images per class:
- als: 2,501
- bean_rust: 2,501
- healthy: 2,286
- unknown: 1,335

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

The original train/test/val split has been preserved in the `split` column.

## Citation

```bibtex
@article{katumba2024leveraging,
  title={Leveraging edge computing and deep learning for the real-time identification of bean plant pathologies},
  author={Katumba, Andrew and Okello, Wayne Steven and Murindanyi, Sudi and Nakatumba-Nabende, Joyce and Bomera, Moses and Mugalu, Ben Wycliff and Acur, Amos},
  journal={Smart Agricultural Technology},
  volume={9},
  pages={100627},
  year={2024},
  publisher={Elsevier}
}
```


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