A Bag of Visual Words Approach for Symbols-Based Coarse-Grained Ancient Coin Classification
This work addresses a domain-specific problem in numismatics for coin classification, but it is incremental as it applies an existing technique with minor modifications.
The paper tackles the problem of classifying ancient coins by issuer using symbols on the coins, proposing a bag of visual words approach with spatial tiling methods, and finds that circular tiling yields the best results.
The field of Numismatics provides the names and descriptions of the symbols minted on the ancient coins. Classification of the ancient coins aims at assigning a given coin to its issuer. Various issuers used various symbols for their coins. We propose to use these symbols for a framework that will coarsely classify the ancient coins. Bag of visual words (BoVWs) is a well established visual recognition technique applied to various problems in computer vision like object and scene recognition. Improvements have been made by incorporating the spatial information to this technique. We apply the BoVWs technique to our problem and use three symbols for coarse-grained classification. We use rectangular tiling, log-polar tiling and circular tiling to incorporate spatial information to BoVWs. Experimental results show that the circular tiling proves superior to the rest of the methods for our problem.