DEVELOPMENT OF A FORM DATA CAPTURE MODEL IN THE 2024 ELECTRONIC ELECTION RECAPITULATION SYSTEM APPLICATION

• Main Author: Chandra, James
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74630
• Institution: Institut Teknologi Bandung

This final project aims to develop the Electronic Recapitulation System (Sirekap) application, which is used as a tool for vote counting, scanning of C-Form election results, and hierarchical vote recapitulation. The final project focuses on the development of the vision engine used as an aid for Optical Character Recognition (OCR) and Optical Mark Recognition (OMR) on C-Form. The current vision engine has several shortcomings that can be further improved. These shortcomings include poor end-to-end model accuracy, user errors in uploading form pages, and a less dynamic vision engine that faces difficulties when sudden changes occur in the C-Form format. This final project takes the opportunity to explore various alternative methods that can be used to address these issues. Some of the explored alternatives include pre-processing image exploration, comparing OCR library experiments with a self-built convolutional neural network OCR, implementing configuration files for engine dynamism, and incorporating page identification features using fiducial markers. The self-built OCR model proves to be the most superior model for digit detection (0-9) with an accuracy of 89% per digit. Additionally, the efforts to achieve dynamism using configuration files and contour detection are considered successful, with an average computation time of only 0.0021 seconds to locate fields on the second page of the C-Form used. The page identification feature is also successfully implemented using the AprilTag fiducial marker, which is praised for its resistance to occlusion, fast computation time, and high reading accuracy.