Visual guided AI color art image generation using enhanced GAN

Read full text Summarise with VIRAT
Authors: Zhuojin Wu
Cited by: 0
References: 30

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

With the improvement of material living standards, the application of Artificial Intelligence technology for generating color art images effectively meets the growing spiritual needs of people. However, traditional manual drawing methods rely on artistic inspiration and require long creative processes, which are inefficient and do not match market demands. This paper puts forward an approach that integrates an Adaptive Attention mechanism and a multi-layer Convolutional Neural Network into the Generative Adversarial Network, and uses Deep Reinforcement Learning to adjust visually guided image information. By dynamically evaluating style loss and color loss between images, the model optimizes the adversarial training process of the generator and discriminator. Experimental results show that the Peak Signal-to-Noise Ratio in the training set increases from 22.5 to 34.8 dB, and the style loss and texture loss stabilize at 0.19 and 0.18 respectively. On the self-built dataset, the Structural Similarity Index Measure of the model reaches 0.52 at the 10th generation, and the average Structural Similarity Index Measure rises to 0.85 after 100 generations of iteration. These results demonstrate that the proposed model can efficiently and automatically generate high-quality images in various artistic styles, providing an innovative solution for the field of artistic painting and meeting the modern demand for personalized and diverse works.

Fields of Study & Subjects

Generative Adversarial Networks and Image Synthesis Digital Media Forensic Detection Advanced Neural Network Applications

Publication details

Journal / source
Scientific Reports
Volume
16
Issue
1
Publication date
19 March 2026
Language
EN