I was doing text recognition from natural scene images using Bag of Words approach with the help of OpenCV library in C++. I can recognize most of the characters from a given scene text image. But to improve prediction accuracy, i thought of adding a simple dictionary kindof stuff at the end, which can predict a feasible word(by making use of a dictionary or so) given a subset of recognized characters from word image. Is it possible to do? recommend a method.
Aucun commentaire:
Enregistrer un commentaire