# ViT Fruit Ripeness Classifier
A fruit ripeness classification model combining **Vision Transformer (ViT)** feature extraction with **Logistic Regression** for fast, accurate inference on CPU or GPU.
## Model Description
This model classifies the ripeness condition of **apples, bananas, and oranges** into three categories:
- **Fresh** - Ready to eat
- **Unripe** - Needs more time to ripen
- **Rotten** - Past optimal consumption
### Architecture
- **Feature Extractor**: ViT Base Patch-16 (`google/vit-base-patch16-224`)
- **Classifier**: Scikit-learn Logistic Regression
- **Feature Dimension**: 768-dim pooled output from ViT
- **Total Classes**: 9 (3 fruits × 3 ripeness states)
## Supported Classes
| Class | Description |
|-------|-------------|
| `freshapples` | Fresh, ready-to-eat apples |
| `freshbanana` | Fresh, ripe bananas |
| `freshoranges` | Fresh, ripe oranges |
| `rottenapples` | Overripe/rotten apples |
| `rottenbanana` | Overripe/rotten bananas |
| `rottenorange…