siddinc / neural_style_transfer

TensorFlow 2.0 Implementation of Neural Style Transfer for creating art using Deep Learning.

Neural Style Transfer

The goal of this project is to implement the Neural Style Transfer (NST) Algorithm by Gatys et al. (2015) using TensorFlow 2.0.

NST is an optimization technique used to take two images: a content image (C) and a style reference image (S) (such as an artwork by a famous painter) and blend them together so the output image (G) looks like the content image, but “painted” in the style of the style reference image.

Tech used:

• TensorFlow 2.0.0
• Python 3.5.6

Pre-Trained Model:

Following the original NST paper, the pre-trained VGG19 ConvNet is used.

This model has already been trained on the very large ImageNet database, and thus has learned to recognize a variety of low level features (at the shallower layers) and high level features (at the deeper layers).

Instructions to run:

• Using anaconda:
  • Run conda create --name <env_name> --file recog.yml
  • Run conda activate <env_name>
• Using pip:
  • Run pip install -r requirements.txt
• Choose a content_image and a style_image from images/samples or take 2 RGB input images of your choice of dimensions approximately near 400x400 pixels
• cd to src
• Run python main.py -cip <path_to_content_image> -sip <path_to_style_image> -noe <no_of_epochs>

Inputs:

Content Image: Octopus

Style Image: Delaunay painting

Output:

Generated Image: