Development of Passport Data Extraction System using Mask RCNN and EasyOCR

The field of document processing has made remarkable strides with the integration of computer vision and machine learning. This progress extends to tasks like text extraction from essential identification documents, such as passports. This process that is excruciating, laborious, and highly prone to errors has created the need for automation. Therefore, this research presents an innovative approach integrating the Mask RCNN Object detection with Optical Character Recognition for the determination of fields of interest in passports including Name, Passport Number, Date of Expiry and many more. Besides, it does not recognize specific formats and layouts, as it is trained on the images of 800 passports of two countries. The system is built into a Flask web application and attains 90.34% accuracy of the reconnaissance and information identification. The results show the ability of deep learning in improving the performance of many travel and security-related tasks, as demonstrated by refinement and adaptability in other areas.