학술논문
Deep Learning Recurrent Attention Optical Character Recognition Network with Data Augmentation for Cheque Data Extraction
Document Type
Conference
Source
2023 International Conference on Computer, Electronics & Electrical Engineering & their Applications (IC2E3) Computer, Electronics & Electrical Engineering & their Applications (IC2E3), 2023 International Conference on. :1-6 Jun, 2023
Subject
Language
Abstract
Bank cheques are often used for monetary transactions in many different fields. Before a check is cashed, it undergoes rigorous scrutiny. The normal verification process includes checking the date, signature, legal information, and cheque amount. This research uses a recurrent attention optical character recognition network (RA_OCRN) to extract legal information from a taken picture of a check in order to address these problems. In this work, we have utilized the recommended deep learning model to do data pretreatment and data augmentation, and we have compared the results to those obtained by using the already available methods. We run the model on the Kaggle cheque detection dataset, and we find that data augmentation improves the algorithm's detection accuracy. Results from the research show that RA_OCRN with data augmentation has the highest recognition accuracy and so has the greatest performance".