Cultural Heritage analysis with YOLO based object detectionBerus, Lucijano (Avtor) Pungerčar, Vesna (Avtor) cultural heritageobject detectionornamentationdeep learningYOLOCultural heritage artefacts that are rich in engraved and embossed ornamentation on vessels, ritual objects, tombstones, and manuscripts. These objects are important for reconstructing social life, ritual practices, and cultural expression a cross regions and periods. To understand a culture, it is not enough to study objects in isolation; systematic comparison across related artefacts is essential to determine whether and how communities were connected. However, such a comparison requires first a robust, scalable detection of their visual content. We therefore study whether a real-time object detection framework can localise and classify ornamentation. In this study, pretrained You Only Look Once version 8 (YOLOv8) and version 11 (YOLOv11) architectures were employed, ranging from their nano to large model versions, to detect ornaments characteristic of Greek and Hallstatt cultural artefacts. YOLOv8 and YOLOv11 were pretrained on Common Objects in Context (COCO) dataset and were able to detect 80different object categories. During the testing of YOLO performance different inherent(YOLO specific) hyper-parameter settings were adopted to detect (localise and classify)ornaments. The models demonstrated promising performance in localising and recognising recurring motifs, yet their accuracy remains constrained by the limited availability of ornament-specific training data. To enhance recognition quality, the development of specialised datasets tailored to cultural ornamentation is essential.20252026-01-29 04:16:15Neznano27215UDK: 004.8:004.93:930.85COBISS_ID: 265308675OceCobissID: 263628291sl