From Elevation Maps To Contour Lines: SVM and Decision Trees to Detect Violin Width Reduction

A team of researchers has developed a novel AI system to automatically detect subtle width reduction in violins, a critical factor in assessing wear and preservation needs for historical instruments. The paper, titled 'From Elevation Maps To Contour Lines: SVM and Decision Trees to Detect Violin Width Reduction,' was authored by Philémon Beghin, Anne-Emmanuelle Ceulemans, and François Glineur and accepted for the Florence Heri-Tech 2026 conference. Their work compares two distinct computer vision approaches applied to detailed 3D photogrammetric meshes of violins.

The first method uses a geometry-based raw representation derived from elevation maps of the 3D scans. The second, more targeted approach relies on feature engineering by fitting parametric contour lines to the violin's shape. The team applied Support Vector Machines (SVM) and Decision Tree classifiers to both data types. Their findings show that while the elevation map approach occasionally delivered strong results, the performance of the contour-based inputs was superior. This indicates that for this specific, nuanced task, a tailored feature extraction process yields more reliable detection than a raw data pipeline.

This research represents a significant application of traditional machine learning (SVM, Decision Trees) to a specialized problem in cultural heritage conservation. The ability to automatically and objectively quantify minute geometric changes from 3D scans provides conservators and luthiers with a powerful diagnostic tool. It moves assessment beyond subjective visual inspection, enabling data-driven tracking of an instrument's condition over time and informing restoration decisions to preserve its acoustic and historical integrity.