Towards Automated Detection of Bookbinding Tool Impressions: A Pilot Study

This paper presents a pilot study on the automated recognition of historical bookbinding tools using deep learning and synthetic image generation. The work focuses on the documented collection of Czech binder Jenda Rajman (1892–1965), whose complete set of metal stamping tools provides an ideal reference dataset. Each tool leaves a characteristic blind-stamped impression on leather bindings, forming a unique visual signature that can support attribution, documentation, and conservation of historical works. To address the lack of complete photographic documentation, two complementary synthetic data generation pipelines were developed: the Procedural Compositional Synthesis and the Background-Normalized Synthetic Stamping methods. These pipelines produce large, annotated image sets that simulate realistic embossing under diverse surface and lighting conditions. A YOLO-based deep learning architecture (YOLOv11 family) was trained and evaluated on datasets generated by both pipelines and their combination. Experimental results demonstrate the feasibility of detecting and classifying small, low-contrast blind-stamped motifs, confirming the potential of synthetic data and deep learning for digital analysis of bookbinding heritage.