Real-Time Neural Style Transfer for Visual Domain Adaptation and Artistic Image Generation Using Deep Learning

Abstract

Neural Style Transfer (NST) is a deep learning technique that generates a new image by combining the semantic content of one image with the visual characteristics of another image’s style. Since the introduction of the convolutional neu-ral network-based style transfer, significant research has been conducted to enhance the quality of stylization, computational efficiency, arbitrary style transfer, content preservation, and style controllability.

However, most of the existing approaches have been evaluated using general artistic datasets and western artistic styles, which are readily recognized. Traditional Indian art forms contain distinctive geometric structures, repetitive motifs, symbolic rep-resentations, decorative patterns, and culturally specific colour combinations. This work proposes a real-time multi-style neural style transfer framework supporting Indian and Western artistic styles using deep CNN feature representations and a curated dataset for custom Indian-art images .