Implementation of OCR and Face Recognition on Mobile Based Voting System Application in Indonesia

Abstract

Elections are a form of democratic practice in Indonesia. Every 5 years an election will be held to elect a president. People who have been able to take part in the election will come to the polling station (TPS) to channel their voting rights. However, this conventional method proved ineffective because some people who were unable to attend due to certain situations for example, traveling out of town, did not want to queue, and experienced illness or physical disability. Therefore, this study aims to design an online voting system based on Android as an alternative to conventional elections and digital transformation in the voting method in Indonesia. The system will use Optical Character Recognition technology by firebase ml-kit to read Identification Number on the Identity Card and face recognition technology to compare the faces of voters during registration and during online elections. The Face Recognition system is implemented using Multi-task Convolutional Neural Network to detect faces and using Tensorflowlite to translate the facial model provided by the FaceNet model. Results Research shows the success of the OCR system is 96.67% and the accuracy of face recognition is 100%. The accuracy of OCR ml-kit and face detection using Multi-task Convolutional Neural Network and Face Recognition using tensorflowlite and FaceNet models proved to be 100% successful.