Deep Learning Model for Text Recognition in Images

Abstract

Computer Vision and its applications are the core of industry digitization which is known as industry 4.0. For automating a process, texts embedded in images are considered as good source of information about that object. Reading text from natural images is still a challenging problem because of complicated background, size and space variations, irregular arrangements of texts. Detection and Recognition are the main stages of reading texts in the wild. In last few years, many researchers have provided many methods for recognizing texts in images. These methods have fine results on horizontal texts only but not on irregular arrangements of texts. This paper mainly focuses on deep learning model for text recognition in images (DL-TRI). The model addresses various cases of curved and perspective fonts and hard to recognize due to complex background.