---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype:
      class_label:
        names:
          '0': angular_leaf_spot
          '1': bean_rust
          '2': healthy
          '3': win32rcparser
  splits:
  - name: train
    num_bytes: 143583487.662
    num_examples: 1034
  - name: validation
    num_bytes: 18492287
    num_examples: 133
  - name: test
    num_bytes: 17700139
    num_examples: 130
  download_size: 179940892
  dataset_size: 179775913.662
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
  - split: test
    path: data/test-*
task_categories:
- image-classification
---


---
dataset_name: bean-disease-uganda
annotations_creators:
  - expert-generated
language_creators:
  - found
task_categories:
  - image-classification
  - computer-vision
task_ids:
  - image-classification
language:
  - en
license: mit
pretty_name: Bean Disease Uganda
tags:
  - agriculture
  - plant-disease
  - imagefolder
  - beans
  - field-data
  - smartphone-images
  - open-access
---

# Bean Disease Uganda Dataset

This dataset contains images of bean leaves categorized into three classes: `angular_leaf_spot`, `bean_rust`, and `healthy`. It is intended for training and evaluating image classification models for plant disease detection.

## Source

The original images were collected by the [Makerere AI Lab](https://github.com/AI-Lab-Makerere/ibean) in collaboration with the [National Crops Resources Research Institute (NaCRRI)](https://www.nacrri.go.ug/). Images were taken in the field using smartphones and annotated by agricultural experts during collection.

## Structure

- `train/`, `validation/`, and `test/` splits
- Each split contains subfolders for the three classes
- Images are in JPEG format, 500x500 pixels

## Usage

```python
from datasets import load_dataset

# Set drop_labels=False to retain the label column
dataset = load_dataset("darcieg/bean-disease-uganda", split="train", drop_labels=False)

image = dataset[0]["image"]
label = dataset[0]["label"]
```

_Note: The `drop_labels=False` flag ensures that the `label` column is retained when loading the dataset. Without it, only the image column will be returned._
## License

MIT License (inherited from the original [ibean GitHub repository](https://github.com/AI-Lab-Makerere/ibean))

## Acknowledgments

This dataset builds on the foundational work of the Makerere AI Lab and NaCRRI. It is intended to make the data more accessible and usable via the Hugging Face Hub.

## Citation

If you use this dataset, please cite:

> Makerere AI Lab. "Bean disease dataset." January 2020. https://github.com/AI-Lab-Makerere/ibean/