Neural style transfer built on TensorFlow and Keras in Python with a deep convolutional neural network (CNN), learning shape of input content and style images. Recombines input images to produce an output resembling content image painted in style of style image.
The technique of neural style transfer is described in the original paper, A Neural Algorithm of Artistic Style by Gatys, Ecker, and Bethge. The goal is to take a content image and a style image, then combine them so that the output image looks like the content image in the style of the style image.
We define two loss functions, the content loss and the style loss. We attempt to transform the output image to minimize both loss functions. Both content and style images are fed into the VGG19 model, and values from the intermediate layers are used to compute the losses.
How to run the application
- Clone the repository.
- Open hyperparameters.py and adjust the hyperparameters as desired.
- Run main.py with
python main.py. The output image is saved in data/output.
- TensorFlow - Deep Learning
This project is licensed under the MIT License - see LICENSE file for details.