Dongba Paintings Multi-Label Emotion Classification Dataset

Based on physical collections preserved in the Lijiang Museum and textual research of relevant literature including Naxi Dongba Paintings, this study systematically collects authentic samples of representative Naxi Dongba paintings, encompassing wooden board paintings, ritual card paintings and scroll paintings. The established sample library covers core cultural scenarios such as traditional prayer & eulogy, nature worship and mythological narration, presenting the Naxi people’s cognitive system of all natural creatures, ritual traditions and social life.Drawing on the research team’s preliminary achievements in cultural semantic parsing, combined with textual research of Naxi cultural classics and historical documents, this study first classifies Dongba painting images into five categories by creative themes: deities, plants, animals, human figures and ghosts. Furthermore, a 14-dimensional emotion classification system is constructed according to the emotional semantics conveyed by the visual symbols of the paintings, which specifically includes approval, bloom, calm, devil, fly, guard, negative, neutral, positive, rest, wilt, work, worried and defensive. The corresponding English labels in the experimental dataset are: approval, bloom, calm, devil, fly, guard, negative, neutral, positive, rest, wilt, work, worried and defensive.The above emotion categories are inherent attributes of the Dongba painting dataset. The emotion label of each image is comprehensively extracted from multi-dimensional visual information, including symbolic elements, painting hues, brushstroke styles, compositional scenes and subject behaviors. This annotation standard demonstrates differentiated adaptability among the five thematic categories.