Style recommendation and simulation for handmade artworks using generative adversarial networks.

Area of Science:

• Computer Science
• Artificial Intelligence
• Digital Art

Background:

• Artificial intelligence (AI) is increasingly utilized in art and design for enhanced creativity.
• Generative Adversarial Networks (GANs) offer new possibilities for artwork creation and style simulation.

Purpose of the Study:

• To present a novel AI model for recommending and simulating handmade artwork styles.
• To improve the accuracy and efficiency of AI-driven artwork generation.

Main Methods:

• A two-phase model employing Generative Adversarial Networks (GANs) for style recommendation and simulation.
• Integration of a genetic algorithm (GA) to optimize Self-Attention (SA) modules for enhanced accuracy.
• Development of the GA-SAGAN model for realistic handmade artwork simulation.

Main Results:

• The GA-SAGAN model significantly reduced validation error in case studies.
• Optimizing SA modules with GA proved effective for generating accurate, realistic artworks.
• The proposed model achieved a near-zero loss rate, with superior Entropy, Precision, and Recall indices compared to baseline GAN and SAGAN models.

Conclusions:

• The GA-SAGAN model demonstrates superior performance in generating diverse and realistic handmade artwork styles.
• Optimizing AI model configurations using genetic algorithms is a viable strategy for improving artistic output.
• AI, particularly GANs enhanced by GA, holds significant potential for revolutionizing digital art creation.