Deep Learning Technique For Transfer Of Artistic Style To Images And Videos
V. V. R. Prasad · i-manager’s Journal on Image Processing · 2020
By composing a complex interplay between the content and style of an image, humanity has mastered the ability to create unique visual experience in fine art, particularly painting. Transfer of the artistic style is a problem in which image style is transformed into image content and generates image stylization. Style transformation can be applied over the entire video sequence by adding image style to video. We use perceptual loss functions to train feed-forward neural network and extract high-level features from the trained networks. We show the effects of image style transfer and video style transfer, through training feed forward network. Our network delivers faster results when compared with Gatys proposed optimization-based method. Resnet is added to the network as an improvisation to transformation network. Pruned Resnet is used, and it gives high computation speed, less size of memory and good performance.