Ancient Bangal Numismatic

Description

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.

Steps to reproduce

1. Collect Ancient Bengal coin images from authenticated private collections and permission-granted auction houses covering the four major traditions: Janapada, Gauda, Harikela, and Samatata. 2. Review all collected images and remove duplicate, damaged, low-quality, or historically uncertain specimens that cannot be reliably classified. 3. Organize the verified images according to the established numismatic classification using standard archaeological and historical references. Group the images into their corresponding traditions, rulers, major types, and final class labels. 4. Preprocess all images using a standardized workflow that includes median filtering, automatic background segmentation, contour refinement, resizing to **512 × 512 pixels**, and conversion to **TIFF (.tiff)** format. 5. Perform manual annotation for each image by recording historical and visual information, including class labels, type names, rulers, motifs, inscriptions, symbols, and source information. Store the annotations in both **CSV** and **XLSX** formats. 6. Verify all annotations using established numismatic literature and expert review to ensure historical accuracy and consistency. 7. Organize the processed images and annotation files into the repository structure and publish the complete dataset through the Mendeley Data repository.

Institutions

• International University of Business Agriculture and Technology

  Dhaka Division, Dhaka

Categories

Licence