FHT: An Unconstraint Farsi Handwritten Text Database

Abstract

Standard databases play very important roles in pattern recognition tasks. To compare the performances of different algorithms, they must be tested on a same dataset. In Farsi, there is not a database of handwritten texts to evaluate different algorithms. In this paper, an unconstraint Farsi handwritten text database is introduced. 250 participants in different ages and education levels filled 1000 forms. Duo to the characteristics of the database, it can be used in many OCR applications. A large number of writers, big lexicon size, out-of-straight textlines and various categories of texts are some of the characteristics.