Speaker
Description
The last kingdom in the Korean peninsula, Joseon dynasty (1392–1897), is known for its meticulous (even obsessive) record-keeping culture. Byeongpung, a form of folding-screen painting, is one such example in art. A byeongpung work, consisted of 6, 8, 10, or 12 panels, is often a cinematic storytelling of a ceremonial or social event — a royal banquet, a civil examination celebration, or a diplomatic reception — rendered with detailed attention to the attire and ranks of participants, architectural settings, ritual utensils, horses and other animals, and even the progression of weather and season across the sequence of panels. The works stand as irreplaceable historical documents as much as aesthetic objects.
Many such paintings were executed on silk (hwa-gyeon) and have suffered deterioration through insect damage, humidity, and molds, particularly among works displaced overseas during the turbulent modern period. A prominent example is the “Welcoming Banquet of the Governor of Pyeongan” (early 1800s; PEM E20262.A-H), which depicts a festivity welcoming civil examination passers in Pyeongan Province. It is now in the collection of the Peabody Essex Museum (USA), and has recently been treated by the Conservation Department of the Leeum Museum of Art — a treatment that required identifying and individually fitting silk infills to approximately 10,000 discrete damage sites by hand.
To address this labor-intensive bottleneck, we developed an interactive machine-learning diagnostic system that automatically detects and digitizes damage sites in high-resolution scans of Korean silk paintings. A conservator selects representative damaged and intact regions, from which a Gaussian Mixture Model is calibrated in the perceptually uniform CIE-LAB color space to classify each pixel and produce binary damage masks for review. The system is deployed as a standalone desktop application that allows conservators to manually correct false positives, and exports per-piece SVG cutting layouts directly scaled for laser-cutting of substrate infills (silk, cotton, or paper) —compressing what was previously a multi-day manual workflow into an automated, auditable pipeline.
We further present a research plan to develop a protocol accelerated aging of substrate materials to be used as physical infills in conservation. While electron beam (EB) irradiation protocols for silk have been practiced in Japan since the 1960s, no equivalent standard exists for cotton or paper substrates. We describe a planned multi-institutional program using electron beams to establish degradation libraries across a dose range of 100–2,500 kGy, with quantitative characterization of mechanical, chemical, and optical properties, alongside structured qualitative evaluation by conservation specialists. This data-driven framework is designed to ultimately inform a decision-support system recommending optimal irradiation conditions given a target substrate age — extending the reach of both AI-assisted diagnostics and radiation processing science into cultural heritage conservation. *This work was supported by the Haorum Alliance Center funded by the Korea Hydro & Nuclear Power Co., Ltd. (KHNP).