An ID Badge Information Extractor Based on Object Detection and Optical Character Recognition
Wallace Gabriel da Silveira Cavalcante, Israel Gondres Torné, Leonardo Camelo, Rubens de Andrade Fernandes, André Luiz Printes, Hendrio Bragança
Abstract
Advancements in Artificial Intelligence and Deep Learning have impacted numerous fields, particularly through innovations like You Only Look Once for object detection and PaddleOCR for optical character recognition in computer vision. These technologies are pivotal in automating and enhancing the accuracy of tasks, such as detecting and extracting characters from identification badges, which have traditionally been prone to manual effort, time consumption, and errors. This study introduces an automated method for detecting and extracting textual data from identification badges, achieving notable efficiency and precision. Our results indicate a Character Error Rate of 0.028 for name recognition and a flawless score for registration number extraction, with a precision rate of 0.992 for the identification badge detection model. By highlighting the importance of automating character extraction and badge detection, this study showcases the ability of Artificial Intelligence and Deep Learning to revolutionize and improve data extraction processes with digital identification systems in professional environments.