Electoral Symbols and Vote detection in Paper Ballots - A Case from Nepal’s Election

Abstract

This paper investigates the potential of advanced object detection technologies to automate and enhance the accuracy and efficiency of the vote counting process in democratic elections that utilize paper-based ballots with electoral symbols. The study focuses on detecting electoral symbols and votes on paper ballots by utilizing two state-of-the-art object detection models: Faster R-CNN, and YOLO. These models were fine-tuned by training them with the ballot papers created using the dataset prepared from the electoral symbols used in Nepal’s general election to ensure high accuracy and reliability in recognizing and validating votes. The system’s effectiveness was demonstrated through a comparison of the models, highlighting the superior performance of Faster R-CNN in terms of precision, despite its slower processing speed compared to YOLO. The results indicate that incorporating object detection technologies into electoral systems can significantly improve efficiency of vote counting process. The study underscores the potential for broader applications of these technologies in promoting transparent and fair elections, especially in countries like Nepal, where traditional paper ballots are still prevalent. This innovative approach ensures a more reliable and efficient electoral process, reducing human error and increasing trust in election outcomes.