Neural Style Transfer:

This code provides a TensorFlow implementation and pretrained models for Artistic Neural Style Transfer, as described in the paper A Neural Algorithm of Artistic Style by Leon A. Gatys, Alexander S. Ecker, Matthias Bethge.

The implementation is supported by Weights and Biases reports. The implementation is divided into two parts:

• Part 1 - Deals with the visualisation of deep embeddings and the content representation of a pre-trained deep learning model.
• Part 2 - Deals with the style representation and the Neural Style Transfer algorithm.

Folder Structure:

.
├── #1Content_Representation.ipynb      - For content representation
├── #2Amalgamation.ipynb                - For amalgamation
├── #3Style_Representation.ipynb        - For style representation
├── #4Style_Transfer.ipynb              - For style transfer
├── README.md
└── Utils                 
    ├── norm_colab.ipynb                - Usage of downloaded and normalizer
    └── vgg-norm.py                     - The normalizer in `tf.keras` 

Data Used:

• ImageNet - For pre-training and normalization

Special Mention: We have used ImageNet dataset downloader to download a specific amount of images from the ImageNet API.

Weights and Models:

• Normalised VGG16 weights
• Unnormalised VGG16 weights

Normalization:

Normalization repository has been used to properly normalize the activation maps of the pre-trained VGG16 model. The code base of the repository has been modified a little bit to be able to harness tf.keras.

Results

• Content Representation

  

• Amalgamation

  

• Style Representation

  

• Neural Style Transfer

  

• Photo-Realistic Images

  

  

Report Authors

Name	Github	Twitter
Aritra Roy Gosthipaty	@ariG23498	@ariG23498
Devjyoti Chakraborty	@cr0wley-zz	@Cr0wley_zz