ArtisticNeuralStyleTransfer

Assignment Certainly! Below is a brief overview of each function and section in Markdown format:

# Neural Style Transfer Readme

## Overview

This repository contains code for performing Neural Style Transfer using PyTorch. Neural Style Transfer is a technique that blends the content of one image with the style of another, creating visually appealing results.

## Setup

### Requirements

- Python 3.x
- PyTorch
- Matplotlib
- Requests
- NumPy

Install the required dependencies using:

```bash
pip install torch matplotlib requests numpy

Usage

1. Load Images:

   • Use the load_image function to load and preprocess images from file paths or URLs.

   content = load_image('/content/golden_gate.jpg').to(device)
   style = load_image('/content/starry_night.jpg', shape=content.shape[-2:]).to(device)

2. Display Images:

   • Utilize the show_image and display_images functions to visualize images.

   content_image = show_image(content)
   style_image = show_image(style)
   display_images(content_image, style_image)

Image Comparison

Content Image

Style Image

3. VGG Model Setup:

   • Load the VGG19 model with pre-trained weights and set to evaluation mode.

   vgg = models.vgg19(pretrained=True).features
   for param in vgg.parameters():
       param.requires_grad_(False)
   vgg.to(device)

4. Feature Extraction:

   • Extract features from content and style images using the VGG model.

   content_features = get_features(content, vgg)
   style_features = get_features(style, vgg)

5. Style Transfer Optimization:

   • Configure parameters and set up the Adam optimizer for optimization.

   show_every = 800
   steps = 40000
   losses = {'content': [], 'style' : [], 'total' : []}
   optimizer = optim.Adam([target], lr=0.003)

6. Optimization Loop:

   • Perform style transfer optimization in a loop, updating the target image.

   for i in range(1, steps+1):
       # ... (Refer to the provided code for the optimization loop)

7. Visualization:

   • Visualize intermediate results at specified intervals.

   if i % show_every == 0:
       print('Iteration {}: Total Loss = {:.2f}'.format(i, total_loss.item()))
       plt.imshow(show_image(target))
       plt.show()

Result Image

Parameters

• Adjust the parameters such as show_every, steps, content_weight, and style_weight based on your specific requirements.

Notes

• Experiment with different content and style images, and tune parameters to achieve desired artistic effects.
• Ensure that the required dependencies are installed before running the code.

Feel free to modify the provided code to suit your specific use case and explore the exciting possibilities of Neural Style Transfer!