AI FOR PRESERVING INDIGENOUS FOLK ART PATTERNS

Abstract [English]

The indigenous folk art traditions represent centuries of cultural wisdom, community belonging, symbolism, and craftsmanship which are specific to a certain region. The continuity of these visual heritage systems is however endangered by fast urbanization, erosion of artisanal transmission and little digital documentation. The presented paper is an AI-based model of the preservation of indigenous folk art patterns by creating systems of data, developing features, and using motifs to reconstruct the object. The suggested approach combines controlled imaging pipelines based on museums, archives, craftsmen and field-based surveys to come up with a very enriched, metadata-conformant, cultural heritage information dataset. Annotation protocols represent the geometry of the motifs, stroke behavior, chromatic palettes, material references and regional semantic attributes that allow the structured cultural representation. The system identifies the discriminative visual features, which are used to define the various folk art traditions, using deep learning models DNNs, ViTs and multimodal style-embedding models. Moreover, encoding symbolic motifs allows comparing styles crossly, and clustering patterns as well as synthesizing them using AI-like synthesis models. Experimental findings indicate that there is a high performance in the motif recognition accuracy, reconstruction fidelity and stylistic consistency indicating a possibility of AI to support artisans, researchers and cultural institutes. The discussion showcases the importance of the ethical aspects of the community involvement, cultural sensitivity and responsible practices of digitization, not just those related to the technical developments. On the whole, the current work adds a scalable, culturally sensitive AI system with the purpose of preserving the indigenous visual knowledge and guaranteeing its continued transfer to the subsequent generations.