This project focus on making the painting of the picture with help of another style image using Neural Style Transfer, some images as on which it is applied are shown below:-
- Introduction
- Technologies
- Launch
- Inspiration
The content and style images will be fed into neural network and their features are extracted then the difference between the generated image,style image and generated image,content image are summed to produce total loss which is then optimized. The generated image features are adjusted according to the total loss.
The languages and libraries used are :-
- Python: 3.7
- Pytorch
- Matplotlib
- Numpy
The pytorch library can be installed using command !pip install torch
and !pip install torchvision
. The code can be executed using python neural_style_transfer.py
The model was inspired by the Coursera Deeplearning.ai specialization and also for more insipartion you can "click_here" and also this "video" .For pytorch tutorials you can prefer this "link"