SURF-2026-0701: Efficient Automated Statistical Recognition of Traditional Stone Window and Door Forms: Integrating AI Computer Vision and H-Scan Interactive System

6000

Traditional stone window and door forms are elements of Suzhou gardens, yet their identification and interpretation remain difficult. This study develops a YOLO- based computer-vision approach for recognition and classification,
integrated with the H-Scan interactive system. It establishes a standardised
annotated dataset, trains and optimises a YOLO model, and applies
the model for recognition, statistical output, and cultural-background
queries. Fourteen predefined categories are used: Square frame
with angled corners, Circular arch, Elongated octagonal, Scepter-
holding, Gourd-shaped, Double-ring, Corner-entry, Han vase III, Flower
vase, Han vase II, Bodhi leaf, Moon- window, Octagonal, and Hexagonal.
The project aims to connect heritage documentation, automated
recognition, and interpretation.