Topic: Neural Style Transfer
The project is a part of Tvarit GmbH Virtual AI-powered Hackathon, held on 28th November 2020.
Underlying Concept: Use of Convolutional Neural Networks for Neural Style Transfer.
Problem Statement: Develop a codebase to perform Neural Style Transfer given a content image and a style image
Concepts used : CNN, VGG-19, PyTorch
We have taken a Content Image and a reference Style Image, the aim is to Stylize the Content Image w.r.t to the Style Image.
Style Weight : 10,000
Content Weight : 1
Optimizer used : LBFGs
For an Example, Below we have taken a content-image (taj.jpg) and a style-image (wave.jpg) and thus a new output-image (stylized.jpg) is generated using both.
Content Image
Style Image
Output Image
