Text detection in images is an active research area which helps to develop a machine/computer automatically read text from images. Model can be made more strong by training it over multiple images and store the result, such that we can use those results in later text recognization and would be easier to recognize them. Challenges will be font characteristics of characters, paper quality, etc.
So the architecture of text recognition will be
(Noise removal → Normalization) → (Segmentation → Feature extraction → Classification) → Store Text data → Output Text data.