Gurmukhi Text: A Dataset for Natural Scene Gurmukhi Text Detection and Recognition

Abstract

Digitization of text plays a vital role in the area of image processing and pattern recognition. However, recognizing text from natural scene has become a challenging task for the researchers due to the challenges of the natural scene along with the lack of benchmark datasets. Public datasets are available for the Latin and Arabic scripts which are useful for research in the field of natural scene text recognition. But for Gurmukhi script no public dataset is available for natural scene text. This paper introduces a new dataset for the natural scene Gurmukhi text images. Dataset contains 500 images of Gurmukhi text which can be used to test the system against different challenges.  Dataset contain complete scene images and can be used for script identification, detection and recognition for Gurmukhi. This paper also provides a survey of benchmark public datasets available for natural scene text recognition.