Interface Homme-Machine pour l'Identification des Liaisons de Coins
Patrice Labedan, Nicolas Drougard
TL;DR
ACCADIL addresses the challenge of linking coin dies by pairing photos from coins with shared characteristics and scoring each pair with a distance metric. The approach combines an image-processing algorithm, based on SSIM-inspired scoring, with an online interactive interface that supports dataset preparation, visual verification, and result export. A key contribution is the end-to-end workflow: type-based data organization, controlled image sizing, and a dashboard-driven workflow that culminates in a downloadable CSV of evaluations and a distance curve with clustering insights, though some visualization features are still under development. The work enables numismatists to efficiently identify likely linked coins within a collection, providing practical, repeatable procedures and user-friendly verification tools, with explicit data-quality and submission constraints to ensure reliable results.
Abstract
ACCADIL is a project that led to the development of software tools for the identification of coin die links from coin photographs. It provides a computational algorithm based on computer vision and classification techniques, along with an online interface for the interactive verification of results. This guide briefly describes the algorithmic principles, the preparation of data prior to analysis, and the features offered by the interface: dataset addition, visualization modes (overlay, side-by-side, magnifier, transparency), result export, and distance visualization. ACCADIL thus provides numismatists with a comprehensive tool for the analysis of die links within a coin collection.