Optical Character Recognition Document

Our purpose is to convert text image data to text and then process the output text to extract some important information. To do that, we have applied some Deep Learning models in Computer Vision to detect the text location on the natural image and then recognize some specific words. We separate our system into multi parts from pre-processing input images to get the final meaning of the text.

As you could see, firstly our system will receive data from the input text image or printed image... This input data will be cleaned or pre-processed by some methods like enhancing the image quality, removing blur, noise, and normalization. Then, the system will run some Deep Learning models to detect the text region on the cleaned input image and recognize, classify each text to some specific word, and at this step, we will have the output text data. Finally, there is an NLP model to clean again this text data to make these text data meaningful and extract the necessary information from them.