Logo Lanfrica

suruleredotdev/symbolic-motif-analysis

Record type:

software
Creator:
sur
Host:
Computational Semiotics, focused on Yoruba artifacts to start # symbolic-motif-analysis A computer-vision pipeline for extracting, vectorizing, and clustering individual carved motifs from photographs of African panel art — Yoruba door panels, Ifa divination boards (opon Ifa), and Benin relief plaques — so that recurring symbolic elements can be found, compared, and labeled across a scattered archival record. --> This is built upon earlier inquiries into African symbolic systems and Yoruba art as historic storytelling, as well as using some of the data aggregation tooling developed as part of our earlier project indexing African artifacts ## Why this exists Carved door panels and divination boards from Yoruba, Ifa, and related West African traditions encode recurring visual vocabularies — figures, knotwork, geometric borders, registers of narrative scenes - which we argue are not merely aesthetic but communicative in function. Other symbolic/writing systems like Egyptian or Mayan hieroglyphs have been the subject of much inquiry and interpretation, using established computational techniques. Our goal is to use such techniques - as well as knowledge of history, mythology and cross-cultural semiotics - to establish an interpretive framework that may be used to expand/enrich the corpus of Yoruba history. As far as the history of such inqury, the Leo Frobenius-led German archeaological expeditions of 1910-1912 as welll as museum collections and archival photo libraries of the Frobenius Institut Bildarchiv, provide a rich set of archaic panels with rich symbolic forms, as well as apparent inquiry into specific motifs and their meaning This pipeline frames "find and cluster the motifs" as a computer-vision problem: segment carved regions out of photographs of varying form and quality (aged B&W photographs, pen-and-ink survey drawings, modern colour photos), normalize them into a comparable visual space, embed them, and cluster the embeddings to surface visual families — then let a human (with LLM assistance) name and …

Languages

Licenses