Neural Style Transfer Project

Project Overview

This project implements neural style transfer, a technique that blends the content of one image with the style of another using a convolutional neural network (CNN). The pre-trained VGG-19 model is used to extract content and style features from the images. By minimizing the content and style losses, a new image is generated that combines the content of the content image and the style of the style image.

Dependencies

The project requires the following libraries and dependencies:

• Python 3.7+
• torch
• torchvision
• Pillow
• numpy
• matplotlib

Code Summary

1. FeatureExtractor Class: Uses the VGG-19 model to extract features.
2. Image Preprocessing: Resizes and normalizes images.
3. Style and Content Loss Calculation: Computes losses to blend style and content.
4. Optimization: Uses the Adam optimizer to update the generated image.

Reference

https://arxiv.org/pdf/1508.06576