Attendscan mobile apps: university student final exam attendance using deep learning

Yii, Vicky Shu Chi (2026) Attendscan mobile apps: university student final exam attendance using deep learning. Final Year Project, UTAR.

PDF Download (5Mb)

Abstract

This project presented AttendScan, an automated attendance marking system for university final examinations that incorporates Deep Learning-based Handwritten Text Recognition (HTR) into a mobile application. In large scale examination venues, manual checking of attendance is normally time-consuming and prone to human error. To address these inefficiencies, the proposed system used a CNN-BiLSTM-CTC architecture to recognise unsegmented and variable length handwritten fields such as index numbers, seat numbers and subject names. This design specifically targets two critical HTR challenges: handwriting variability and limited character scope. To provide strong recognition of various writing styles and alphanumeric characters, the model was trained on a hybrid dataset that included real exam attendance slips and publicly available handwriting datasets. To obtain optimal performance, the system employed Optuna to automatically optimise hyperparameters, which significantly reduced Character Error Rate and Word Error rate, and improved Exact Match Accuracy. To illustrate the practical application of HTR towards automation, the model was implemented as a RESTful API to be called by mobile applications. A student-facing mobile application can be used to capture attendance slips, send them to the API, and submit the recognised results to mark attendance. Meanwhile, a dedicated web platform allows administrators to manage student attendance records. This project combines the deep learning recognition with the practical use and thereby reduces the administrative burden and offers a scalable solution to the modern education institutions.

Actions (login required)