Image Style Transfer

Implementation of "Image Style Transfer Using Convolutional Neural Networks" (Gatys et al.)

---

Overview

This project implements Image Style Transfer, a technique that combines the content of one image with the style of another image to generate a new artistic output.

• Implemented from scratch using PyTorch, without relying on pre-built style transfer libraries.

• Includes key components such as:

  • Content loss
  • Style loss (Gram matrices)
  • Total variation loss
  • Optimization with L-BFGS

For a detailed explanation of the project, concepts, and my learning journey, check out my blog post: Here

---

✨ Features

• PyTorch implementation of Neural Style Transfer.
• Optimized using L-BFGS optimizer for faster convergence.
• Adjustable weights (alpha for content, beta for style).
• Extendable codebase for experiments with new loss functions or architectures.

---

Installation

# Clone repository
git clone https://github.com/nirdesh17/style-transfer.git
cd style-transfer

---

Usage

Run style transfer with your own content and style images:

python src/run.py --content path/to/content.jpg --style path/to/style.jpg --output output.jpg

Optional changes:

• iterations : Number of optimization steps (default: 500)
• alpha : Content weight (default: 1)
• beta : Style weight (default: 1e6)

---

Results

Content Image	Style Image	Output

Project Structure

style-transfer/
│── src/                
│   ├── run.py          
│── images/            
│   ├── content
│   ├── style
│   ├── outputs
│── README.md           

---

Future Work

• Multi-style transfer (combine multiple styles).
• Real-time transfer using feed-forward networks.
• Web/GUI demo for interactive use.

---

Acknowledgements

• Image Style Transfer Using CNNs (Gatys et al.)
• PyTorch Documentation
• NumPy Documentation

---

Connect with Me

• LinkedIn

---