Neural Style Transfer Using CNN

Neural Style Transfer (NST) is a technique which transfers one of two images into the other.

Combing two images which are content image and style image. Extracting the style from style

image and applying it on the content image as it used as object to change the content image

appearance. In this project, Neural Style Transfer (NST) is a model that can modify the

appearance of the image style. The NST consists of two parts which are content image and style

image. The used architectures for this model are mostly CNN. Neural style transfer employs a

pre-trained Convolutional Neural Network for feature extraction and separation of content and

style representations from an image. Neural style transfer network has two inputs: Content

image and Style image. The content image is recreated as a newly generated image which is the

only trainable variable in the neural network. The architecture of the model performs the

training using two loss terms: Content Loss and Style Loss. Content Loss: by applying the mean

square MSE difference between matrices generated by the content layer, when passing the

generated image and the original image. Style Loss: calculating the gram matrix. The gram

matrices calculation involves calculating the inner product between the vectorized feature

maps of a particular layer. The Gram matrix represents the inner product of each vector and its

corresponding vectors within the same matrix. Its significance in contemporary machine learning

stems from applications in deep learning loss, particularly in the computation of loss functions during

style transfer, which relies on the gram matrix.