Ancient Bangal Numismatic

This dataset presents a comprehensive collection of 2,704 annotated images of Ancient Bengal coins spanning approximately the 2nd century BC to the 10th century AD. The collection covers four major historical coin traditions of Ancient Bengal: Janapada, Gauda, Harikela, and Samatata, representing the evolution of Bengal's monetary system over nearly twelve centuries. The dataset is organized into 52 machine learning-ready classes, including 3 Janapada classes, 6 Gauda classes, 6 Harikela classes, and 37 Samatata classes derived from the standard 22-type historical classification. The images were collected from five permission-granted international auction houses and authenticated private collections. Each coin image was manually verified and annotated using established archaeological and numismatic references. The annotation process records important historical and visual information, including coin type, class, ruler, inscriptions, motifs, symbols, and other distinguishing features. Annotation files are provided in both CSV and XLSX formats, allowing straightforward integration into machine learning, computer vision, and digital heritage research workflows. To ensure consistency, all images were processed using a standardized preprocessing pipeline that included noise reduction, automatic background segmentation, contour refinement, resizing to 512 × 512 pixels, and conversion to TIFF (.tiff) format. The preprocessing workflow preserves the original appearance and historical characteristics of the coins without applying artificial enhancement or data augmentation. The repository follows a hierarchical directory structure consisting of separate image folders for each historical tradition together with corresponding annotation files. This organization allows researchers to perform classification at multiple levels, including tradition, ruler, major type, or fine-grained class, depending on the objectives of their study. The dataset provides a valuable resource for computer vision, machine learning, archaeology, numismatics, digital humanities, cultural heritage preservation, image retrieval, feature extraction, fine-grained image classification, similarity search, and historical coin analysis. It also supports the development and benchmarking of artificial intelligence methods for ancient coin recognition while preserving the rich historical and cultural heritage of Ancient Bengal.