---
library_name: transformers
license: apache-2.0
base_model: timm/mobilenetv3_small_100.lamb_in1k
tags:
- image-classification
- vision
- timm
- generated_from_trainer
datasets:
- AI-Lab-Makerere/beans
metrics:
- accuracy
pipeline_tag: image-classification
model-index:
- name: beans-image-classifier
  results:
  - task:
      type: image-classification
      name: Image Classification
    dataset:
      name: AI-Lab-Makerere/beans
      type: AI-Lab-Makerere/beans
    metrics:
      - type: accuracy
        value: 0.9549
        name: Validation Accuracy
      - type: accuracy
        value: 0.8984
        name: Test Accuracy
---

# Bean Leaf Disease Classifier

Fine-tuned [MobileNetV3-Small](https://huggingface.co/timm/mobilenetv3_small_100.lamb_in1k) on [AI-Lab-Makerere/beans](https://huggingface.co/datasets/AI-Lab-Makerere/beans).

Given a photo of a bean leaf, the model predicts one of:

- `angular_leaf_spot`
- `bean_rust`
- `healthy`

## Metrics

| Split | Accuracy | Loss |
| --- | ---: | ---: |
| Validation (best checkpoint) | 95.49% | 0.235 |
| Test | 89.84% | 0.303 |

Training ran for 5 epochs on Apple Silicon (MPS). The best checkpoint was restored before export.

## Usage

```python
from transformers import pipeline

clf = pipeline("image-classification", model="omaralshareef/beans-image-classifier")
print(clf("leaf.jpg"))
```

## Training hyperparameters

- learning rate: `5e-5`
- train batch size: 16
- eval batch size: 16
- epochs: 5
- optimizer: AdamW
- seed: 42

## Training results

| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
| 0.5874        | 1.0   | 65   | 0.4124          | 0.8346   |
| 0.4042        | 2.0   | 130  | 0.2352          | 0.9549   |
| 0.2805        | 3.0   | 195  | 0.1679          | 0.9549   |
| 0.2367        | 4.0   | 260  | 0.1370          | 0.9549   |
| 0.1933        | 5.0   | 325  | 0.1394          | 0.9398   |