Neural-Style-Transfer-using CNN

This repo contains the jupyter notebook describing the implementation of Neural Style Transfer using Convolution Neural Networks

Neural Style Transfer is transfering a picture's style to another.

• Its implementation is used in the android app PRISMA

It is done using a pre-trained VGG19 model which is equivalent to transfer learning. All the pretrained models are trained on GPUs

• This is the link to all the pretrained models available.
• Download the model and save it in the folder where jupyter notebook file is present.

The input and outputs are in their respective folders. The model ran on few iterations so the output might not be impressive. The nst_utils.py contains necessary functions to resize, normalize and to generate noise image.

NOTE:

Use GPU if its installed on your PC or use cloud services like Google Colaboratory etc to carry out the model for large number iterations say 50000 for best results.