Gupta Coin Grading and Rarity Dataset

Description

The Gupta Coin Grading and Rarity Dataset is a manually annotated primary dataset developed to support research on automatic grading and rarity analysis of ancient Gupta coins. The dataset contains high-quality coin images collected from reliable numismatic sources and is organized to facilitate computer vision and machine learning applications. The dataset is distributed into three main components: Images, Metadata, and Mapping Files. The image collection is organized into four metal categories, namely gold, silver, copper, and lead. Within each metal category, the images are further arranged according to the corresponding Gupta ruler, allowing users to access both metal-wise and ruler-wise coin collections. Each folder contains the original coin images identified by a unique image filename. The metadata component contains four CSV annotation files corresponding to the four metal categories. Each metadata record is linked to a unique coin image through the coin_id field and provides manually verified annotation information. The metadata includes the normalized coin grade, rarity information, and, where available, additional provenance information such as the source website, source URL, and license status. This organization enables direct integration of image data with structured annotations for machine learning, retrieval, and digital heritage applications. To ensure consistent grading across the dataset, a separate mapping file (grade_map.csv) is provided. This file standardizes the relationship between the original grading terminology, abbreviated grade labels, normalized grade labels, and their corresponding numerical grading ranges. The mapping file simplifies label normalization and supports reproducible preprocessing across different grading systems. Overall, the dataset contains 2,557 manually annotated Gupta coin images, including 2,063 gold, 303 silver, 146 lead, and 45 copper coins. The hierarchical folder organization, comprehensive metadata annotations, and standardized grade mapping provide a structured benchmark for automatic coin grading, rarity prediction, historical coin analysis, and other computer vision tasks related to ancient numismatics.

Institutions

• International University of Business Agriculture and Technology

  Dhaka Division, Dhaka

Categories

Licence