Authors : Mouad Tali; Mesut Cevik

Volume/Issue : Volume 10 - 2025, Issue 4 - April

Google Scholar : https://tinyurl.com/y6p8br7d

Scribd : https://tinyurl.com/yjbfypuy

DOI : https://doi.org/10.38124/ijisrt/25apr1666

Google Scholar

Note : A published paper may take 4-5 working days from the publication date to appear in PlumX Metrics, Semantic Scholar, and ResearchGate.

Note : Google Scholar may take 15 to 20 days to display the article.

Abstract : Generative Adversarial Networks (GANs) represent a groundbreaking advancement in computational creativity, enabling machines to synthesize art, music, and literature with unprecedented realism. This study critically evaluates the technical and ethical dimensions of GANs in artistic contexts, focusing on StyleGAN’s performance on the WikiArt-27K dataset, a comprehensive repository spanning 27 diverse artistic styles from Baroque to Cubism. Through rigorous experimentation, we demonstrate that StyleGAN achieves a Fréchet Inception Distance (FID) score of 15.3, approaching the perceptual quality of human-created art (10.8). However, persistent technical challenges such as mode collapse— observed in 30% of trials, where generators produce repetitive outputs—and high-resolution artifacts (e.g., blurred textures and color banding at resolutions exceeding 2048x2048 pixels) hinder practical adoption. Qualitative surveys of 50 professional artists and critics reveal a 23% preference for human-AI collaborative artworks, underscoring hybrid creativity’s potential to democratize artistic expression and bridge the gap between human intuition and algorithmic precision. To address ethical concerns, we propose actionable frameworks, including dual attribution protocols to resolve authorship disputes and adversarial debiasing techniques to mitigate cultural bias in training datasets. By advocating for transparency through blockchain-based metadata and standardized disclosure labels, this work positions GANs as tools to augment—not replace—human creativity, fostering interdisciplinary collaboration between artists, technologists, and policymakers. Our findings highlight the urgent need for ethical guidelines and technical innovations to ensure AI- generated art aligns with societal values while expanding creative possibilities.

Keywords : Generative Adversarial Networks (GAN), Ethical Guidelines, Cultural Bias, Human-AI Collaboration, Computational Creativity.

References :

1. Goodfellow, I., et al. (2014). Generative Adversarial Networks. NeurIPS.
2. Chen, Z., & Zhang, Y. (2024). CA-GAN: Synthesis of Chinese Art Paintings. The Visual Computer.
3. Elgammal, A., et al. (2021). The Ethics of AI Art. Frontiers in Digital Humanities.
4. Gao, M., & Pu, P. (2024). GANs in Digital Media Arts. IEEE Access.
5. Karras, T., et al. (2021). StyleGAN3: Alias-Free Generative Adversarial Networks. CVPR.