The goal of Scene Text Recognition is to properly localize and transcribe chunks of text which appear in photographs. Until recently, it has been exclusively problem of computer vision. Individual character were detected in the image and then glued together to form words.
The goal of this project is to go beyond this state and use language knowledge to improve the recognition perfomance. The additional goal is to combine the visual and language knowledge to implicitly model the text in the photograph relate to the scene it is in.