Design of a Deep Learning Model for Bronze Dagger Morphology Classification

Abstract. In archaeology, buried cultural artifacts serve as important material evidence for distinguishing historical periods, and inferring the morphology and chronology of artifacts excavated from archaeological sites is a crucial task. Buried cultural artifacts can be used as fundamental data for inferring the characteristics of archaeological sites and the scale of past groups that utilized the sites through their morphology and chronological context. In particular, bronze dagger, which are among the buried cultural artifacts excavated from prehistoric sites, exhibit different morphologies according to their periods, making the morphological classification of bronze dagger a key indicator for determining the chronology of archaeological sites. However, the various forms of bronze dagger excavated from the Korean Peninsula to northeastern China show limitations for manual classification by archaeological researchers, and the ambiguous characteristics where forms are not clearly distinguishable create problems in securing consistent objectivity in bronze dagger chronological classification. To overcome these limitations, this paper proposes a framework for automatically classifying bronze dagger morphology using deep learning-based image classification models and quantitative results for bronze dagger classification tasks.