lelinquan-YAIG-mini

Launched in 2025, YAIG-mini is an academic dataset dedicated to the visual understanding of Chinese classical garden culture. It contains a total of 11,804 real-scene images aggregated from tourist platform on-site photographs, archived ancient garden illustrations and professionally field-shot pictures, covering landmark classical gardens including the Humble Administrator’s Garden, Lingering Garden and Summer Palace. Adopting a three-tier structured annotation scheme, the dataset defines six macroscopic garden scene categories at Level 1 and subdivides into 35 refined gardening components at Level 2. Annotation quality is guaranteed via a three-stage workflow consisting of model-assisted pre-labeling, double-blind cross-check and final review by professional garden scholars. Supporting diverse research tasks such as multi-level garden image classification and cultural artistic conception retrieval, Lelinquan fills the gap of datasets focusing on cultural semantics for classical gardens and facilitates academic researches on digital revitalization of gardens, intelligent retrieval of cultural heritage and intelligent visual analysis of ancient architecture.