---
pipeline_tag: image-classification
datasets:
- AI-Lab-Makerere/beans
tags:
- image-classification
- pytorch
- transformers
---

# Bean Leaf Disease Classifier

Fine-tuned from `facebook/deit-tiny-patch16-224` using the Beans dataset.

## Classes
angular_leaf_spot, bean_rust, healthy

## Training
- 600 training images
- 2 epochs
- Two learning-rate experiments tracked with MLflow
- Best model selected using validation accuracy

## Evaluation
- Validation accuracy: 0.9549
- Test accuracy: 0.9297
- Test macro F1: 0.9302

## Limitations
This is an educational model evaluated on a small dataset.
It is not validated for operational agricultural diagnosis.