Visual guided AI color art image generation using enhanced GAN.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Digital Art Generation

Background:

• Traditional art creation is time-consuming and relies on inspiration, failing to meet market demands.
• Growing demand for personalized and diverse digital art necessitates efficient generation methods.

Purpose of the Study:

• To develop an AI-driven approach for automated, high-quality color art image generation.
• To integrate advanced AI techniques for optimizing artistic image synthesis.

Main Methods:

• An approach combining Generative Adversarial Networks (GANs) with an Adaptive Attention mechanism and multi-layer Convolutional Neural Networks.
• Deep Reinforcement Learning was employed to fine-tune visually guided image information.
• Dynamic evaluation of style and color loss optimized the adversarial training process.

Main Results:

• Peak Signal-to-Noise Ratio improved from 22.5 to 34.8 dB.
• Style and texture losses stabilized at 0.19 and 0.18, respectively.
• Structural Similarity Index Measure reached 0.52 at the 10th generation, averaging 0.85 after 100 generations.

Conclusions:

• The proposed model efficiently generates high-quality images in diverse artistic styles.
• This AI approach offers an innovative solution for artistic painting, meeting modern demands for personalized visual content.