An AI-Based Framework for Automated 3D Pattern Generation and Virtual Fitting of National Garments

Backgraund: This research aims to develop a fully automated 3D pattern generation and virtual fitting system for national garments based on artificial intelligence and machine learning technologies. Methods: The study collected 10,247 images of national garments and decorative elements. Convolutional Neural Networks (ResNet-50, EfficientNet-B7) and Vision Transformer (ViT) architectures were trained for decorative element recognition. Generative Adversarial Networks (StyleGAN2) were applied for 2D-to-3D pattern conversion. Point Cloud Library and Marching Cubes algorithm were used for 3D body scan data processing. Physics-based cloth simulation was implemented in Unity3D environment. Results: The developed AI model recognized decorative elements with 96.3% accuracy. The 2D-3D conversion algorithm achieved 87.5% geometric precision. The virtual fitting system demonstrated 92.8% correspondence with real garments. The system reduced production time by 2.7 times and decreased design errors by 68 Conclusion: The AI-based automated system significantly increases efficiency in national garment production, reduces costs, and enables creation of personalized products.