The visual communication using generative artificial intelligence in the context of new media

Area of Science:

• Artificial Intelligence
• Computer Vision
• New Media Studies

Background:

• Generative Artificial Intelligence (AI) offers novel methods for visual communication.
• Existing models face challenges in generating high-fidelity images that meet specific conditions.
• The integration of advanced AI architectures is crucial for progress in new media content creation.

Purpose of the Study:

• To explore visual communication methods using generative AI.
• To propose and evaluate an image automatic generation and recognition model.
• To enhance image quality and recognition accuracy for new media applications.

Main Methods:

• Developed a model integrating Conditional Generative Adversarial Network (CGAN) with the Transformer algorithm.
• Utilized noise vectors and conditional variables as generator inputs.
• Incorporated a Transformer module with multi-head self-attention for complex relationship modeling and long-range dependency capture.

Main Results:

• Achieved an accuracy of 95.69%, surpassing the baseline Generative Adversarial Network by over 4%.
• Attained a Peak Signal-to-Noise Ratio of 33dB, indicating superior image quality.
• Recorded a Structural Similarity Index of 0.83, confirming enhanced image generation and recognition.

Conclusions:

• The proposed model demonstrates high recognition and prediction accuracy for generated images.
• The model achieves superior image quality, essential for visual communication in the new media era.
• This AI-driven approach holds significant application value for future visual content creation and analysis.