---
license: mit
language: en
tags:
  - steganalysis
  - image-classification
  - tensorflow
  - transformer
model-index:
  - name: Stegformer
    results: []
---

# 🕵️‍♀️ Stegformer - Model for Steganalysis

This repository contains a TensorFlow-trained model saved in `saved_model` format, designed to detect images manipulated through steganographic techniques. The model uses a Transformer-based architecture and is optimized for grayscale image inputs of size 256x256.

---

## Model Details

- **Expected input**: Grayscale image of shape `(256, 256, 1)`
- **Task**: Binary classification (`stego` vs `cover`)
- **Framework**: TensorFlow 2.x
- **Format**: `saved_model`

---

## Usage

You can download and load this model in any environment that supports TensorFlow.

### Requirements

```bash
pip install tensorflow huggingface_hub
```

---

### Download and Load

```python
from huggingface_hub import snapshot_download
import tensorflow as tf

# Download the model from Hugging Face Hub
model_path = snapshot_download(repo_id="MarilineDelgado/stegformer")

# Load the model in SavedModel format
model = tf.keras.models.load_model(model_path)
```

---

### Example Prediction

```python
import numpy as np

# Test image (1, 256, 256, 1) - grayscale
image = np.random.rand(1, 256, 256, 1).astype(np.float32)

# Make prediction
prediction = model.predict(image)

print("Prediction:", prediction)
```

---

## License

This model is licensed under the [MIT License](https://opensource.org/licenses/MIT). You are free to use and modify it for academic or personal purposes.

---

## Contact

Created by  [MarilineDelgado](https://huggingface.co/MarilineDelgado).  
For questions or collaboration, please open an issue or contact me through the platform.