Style Transfer Using Convolutional Neural Network and Image Segmentation
Minyeo Kim, Hyeong-Seok Choi, Joonki Paik · TECHART Journal of Arts and Imaging Science · 2021
In this paper, we present a style transfer-based artwork using deep learning. Our proposed artwork aims to perform a different style transfer to the foreground and background using images of different styles. The framework of our proposed artwork comprises three steps: i) human detection and segmentation, ii) style transfer to the human and background regions, and iii) fusion of the style-transferred result images. The resultant style transferred image is preserved the content of the original image while changing the original style with the target style. The proposed artwork can provide a different artistic effect and style with an esthetic sense to human and background regions using different style images.