A Study of Deep Convolutional Generative Adversarial Networks for Style Transformation and Generation in Online Artwork Creation

Abstract

Deep Convolutional Generative Adversarial Networks (DCGAN) bring new opportunities for online artwork creation in the context of rapid development of digitalisation and artificial intelligence technologies. This study analyses its application in art creation, discusses the principles of Generative Adversarial Networks (GANs) and Deep Convolutional Generative Adversarial Networks (DCGANs), and describes the architecture of DCGANs and their advantages in image generation. We analyse online art creation platforms, gain insights into the needs of creators, and reveal the limitations of existing technologies. On this basis, the DCGAN-based art style conversion algorithm is successfully implemented, which can effectively extract and migrate style features to generate works with visual effects similar to the target style and retain key information of the original content. Meanwhile, the artwork generation model constructed based on DCGAN has a powerful generative capability, which can generate artworks based on random noise or text descriptions, and achieve diversity and controllability by adjusting parameters and input conditions. The research results provide theoretical support and technical reference for online art creation and promote the development of this field.