Handwritten Text Recognition Algorithm

This work emerged as a proposed solution to a pressing need detected within PROCER, an Argentinian company dedicated to the development of technological solutions for visually impaired people. Through surveys and users’ feedback, it became evident that the implementation of handwritten text reading was a need among users of the company’s devices. In order to develop an algorithm for the detection of handwritten text in Spanish that improves the quality of life of these people, the authors sought to provide a solution from Artificial Intelligence. To achieve this goal, two Deep Learning neural network models were implemented: inductive transfer learning with You Only Look Once (YOLO) and Convolutional Neural Networks (CNN). In addition, Optical Character Recognition tools were incorporated to complement these models. The culmination of this project is a functional algorithm prototype that achieves a digitized text of the original manuscript with a minimum Levenshtein distance of 17 points from the normalized reference text.