Databases of mathematical documents
Masakazu Suzuki, Christopher D. Malon, Seiichi Uchida · QIR (Kyushu University Institutional Repository) (Kyushu University) · 2007
This paper describes the specifications for three ground-truthed mathematical character and symbol image databases, called InftyCDB-1, InftyCDB-2, and InftyCDB-3. In the former two databases, the ground-truth of each character is composed of type, font, quality (touching/broken) and link (relative position), etc. InftyCDB-1 includes all the characters and symbols of 30 articles on mathematics, and is organized so that it can be used as word image database or as mathematical formula image database. InftyCDB-2, which is a continuation of InftyCDB-1, includes 37 articles including French and German articles and is organized like InftyCDB-1. InftyCDB-3 is a single character database for training and evaluating single-character recognition engines.