---
license: apache-2.0
pipeline_tag: image-classification
tags:
- model_hub_mixin
- pytorch_model_hub_mixin
- medical-imaging
- cervical-cancer
- pap-smear
- pytorch
- alexnet
---



# AlexNet Fine-Tuned on HERLEV Dataset

This repository contains a fine-tuned **AlexNet** model trained on the **HERLEV cervical cytology dataset** for multi-class classification of Pap smear images.

The model was uploaded using the **PyTorchModelHubMixin**, enabling native Hugging Face loading via `from_pretrained`.

---

## Model Details

- **Architecture:** AlexNet  
- **Framework:** PyTorch  
- **Input size:** 224 × 224 RGB  
- **Number of classes:** 7  
- **Task:** Cervical cell image classification  

---

## Classes

| Label | Cell Type |
|------|-----------|
| 0 | Superficial Squamous |
| 1 | Intermediate Squamous |
| 2 | Columnar |
| 3 | Mild Dysplasia |
| 4 | Moderate Dysplasia |
| 5 | Severe Dysplasia |
| 6 | Carcinoma in situ |

---

## How to Use

### Installation


```bash
pip install torch torchvision huggingface_hub pillow
```

---

## Load the Model

```python
import torch
from huggingface_hub import PyTorchModelHubMixin
from torchvision import transforms
from PIL import Image

from model import AlexNetHERLEV

model = AlexNetHERLEV.from_pretrained(
    "hp1318/alexnet-finetuned-herlev"
)

model.eval()
```

---

## Image Preprocessing

```python
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    )
])
```

---

## Run Inference

```python
image = Image.open("cell_image.jpg").convert("RGB")
image = transform(image).unsqueeze(0)

with torch.no_grad():
    outputs = model(image)
    prediction = torch.argmax(outputs, dim=1)

print("Predicted class:", prediction.item())
```

---

## Output

The model outputs logits for all cervical cell classes.  
The predicted label corresponds to the class with the highest logit score.

---

## Notes

- Input images must be RGB format.
- Images are resized to 224 × 224 before inference.
- Normalization follows ImageNet statistics.
- This model is intended for research and educational use only.

---

## Citation

If you use this model in academic work, please cite the corresponding paper or repository.