Neural Style Transfer Web App

A beautiful and interactive Streamlit web application that applies Arbitrary Neural Style Transfer to any uploaded image using the pre-trained Magenta model from TensorFlow Hub.

Transform your photos into stunning artworks inspired by 9 different artistic styles in seconds — powered entirely by deep learning!

Live Demo

Run it locally and experience the magic!

Features

• Upload any image (PNG, JPG, JPEG)
• Choose from 9 unique pre-defined artistic styles
• Real-time style transfer using Google's Magenta Arbitrary Image Stylization model
• Side-by-side comparison: Original → Style → Stylized Result
• High-resolution downloadable output (1024×1024 PNG)
• Clean, responsive, and user-friendly interface
• Fast inference with cached model and image loading

Screenshots

(Add your own screenshots later by replacing this section)

Repository Structure

Neural-Style-Transfer/
├── nst.py                     # Main Streamlit application
├── requirements.txt           # Python dependencies
├── README.md                  # This file
├── Style 1.jpg                # Pre-defined style images (1 to 9)
├── Style 2.jpg
├── ...
└── Style 9.jpg

Requirements

• Python 3.8+
• Streamlit
• TensorFlow (2.x)
• TensorFlow Hub
• Pillow (PIL)
• NumPy

Install all dependencies easily with:

pip install -r requirements.txt

requirements.txt content (for reference)

streamlit
tensorflow
tensorflow-hub
Pillow
numpy

How to Run

1. Clone or download this repository
2. Navigate to the project folder:

   cd Neural-Style-Transfer

3. Install dependencies:

   pip install -r requirements.txt

4. Run the app:

   streamlit run nst.py

5. Open your browser (usually at http://localhost:8501)
6. Upload an image, select a style, and watch the AI create art!

How It Works

• Uses the official Magenta Arbitrary Image Stylization v1-256 model from TensorFlow Hub
• Automatically crops input to square, resizes to 256×256 for optimal processing
• Applies fast style transfer combining content and selected style image
• Upscales final result to 1024×1024 for high-quality downloads

Credits

• Model: Magenta Team @ Google
• Neural Style Transfer Research: Gatys et al., and fast arbitrary stylization by Google
• Built with ❤️ using Streamlit and TensorFlow

Author

Agaba_Embedded
Transforming pixels into art, one upload at a time.

---

Enjoy creating AI art! 🖼✨
Feel free to fork, star ⭐, and customize with your own styles!