Pytorch Implementation Of Neural Style Transfer Utilizing VGG19 Architecture

Neural Style Transfer is an optimization technique which take two images as input generally called content and style image, then it tries to generate a new image having content (objects) of the content image and style (colors) of the style image.

In this repository I've used VGG19 for extracting content and style of the images. Content and Style are extracted from different convolutional layer of VGG19.

• conv1_1
• conv2_1
• conv3_1
• conv4_1
• conv5_1

• conv4_2

• Python 2.7
• Pytorch
• Cuda

• Clone the repository
• Load the content and style image (change the path)
  content=loadImage('images/23.png').cuda()
style = loadImage('images/2.jpg', shape=content.shape[-2:]).cuda()
• Run the code using python styletransfer_gc.py. Output will be in images folder
• If in output you're not noticing enough style try increasing the style_weight. You can also increasing steps.

Content Image	Style Image	Output at Different Steps