Deep learning-based intelligent generation model for intangible cultural heritage products

Abstract

Rapid digital transformation highlights the urgent need to preserve Intangible Cultural Heritage (ICH), as traditional methods face declining craftsmanship, limited innovation, weak cultural conditioning, poor optimization, and reduced creative diversity, leading to less authentic and visually inconsistent designs. To overcome these limitations, this research suggests a Deep Learning-Based Intelligent Generation Model integrating an Intelligent Penguins Search–driven Conditional Generative Adversarial Network (IPS-ConGAN), forming a novel synergy between deep generative learning and evolutionary optimization for ICH product design. The proposed framework is evaluated using a Chinese intangible cultural heritage image dataset containing paintings, motifs, calligraphy, and regional patterns, where each sample includes category and area annotations. The dataset contains approximately 119,000 ICH image files. Data preprocessing involves image resizing and Gaussian filtering to ensure consistent input quality and reliable feature learning. CNN is employed for feature extraction to capture deep semantic, spatial, and structural characteristics of traditional motifs, architectural elements, and historical patterns, enabling accurate representation of cultural essence. ConGAN facilitates culturally aware feature learning and conditional generation, while IPS refines generated designs to enhance visual harmony, creativity, and convergence efficiency. The model is implemented using Python-based tools and experimentally validated. Results demonstrate superior performance with 94.56% accuracy, 95.21% precision, 93.35% recall, 95.62% F1-score, 96.5% AUC, and 0.0012 MSE. Compared to traditional GAN, the proposed method achieves a higher accuracy improvement of 11.25%, demonstrating superior performance in ICH product generation and classification. Overall, the proposed framework effectively bridges cultural preservation and digital innovation for creative industries.