ARCHIVES
VOL. 8, ISSUE 1 (2024)
Optimized dataset creation for real time facial recognition
Authors
Satisfy Ogunjebe, Uchendu Iyemeh
Abstract
This study investigates optimizing dataset
creation for real time facial recognition attendance tracking. It explores the
impact of grayscale conversion intensity values on preprocessing time and
accuracy. Data from six staff members, with 80 images each, were collected
under controlled conditions. Grayscale conversion was applied with intensity
values of 255, 127, and 95, followed by face detection and feature extraction.
SVM and KNN classification algorithms were assessed across the three datasets,
with the 127 intensity value showing the most promising balance between time
and accuracy. SVM achieved accuracies of 96.88%, 94.79%, and 94.79% with
preprocessing times of 0.036, 0.02, and 0.017 seconds for intensity values 255,
127, and 95, respectively. KNN accuracies were 88.54%, 87.50%, and 81.25%, with
preprocessing times of 0.036, 0.02, and 0.017 seconds for intensity values 255,
127, and 95, respectively. Lower intensity values like 127 and 95 demonstrated
efficiency in balancing accuracy and preprocessing time, aiding in efficient
attendance tracking. Grayscale conversion value of 127 offers a compelling
balance between accuracy and efficiency, recommended for optimized facial
recognition datasets in real-time attendance tracking applications. The study's
significance was highlighted through comparisons with existing literature,
positioning it as a pioneering effort in dataset optimization for real-world
applications.
Download
Pages:8-15
How to cite this article:
Satisfy Ogunjebe, Uchendu Iyemeh "Optimized dataset creation for real time facial recognition". International Journal of Advanced Engineering and Technology, Vol 8, Issue 1, 2024, Pages 8-15
Download Author Certificate
Please enter the email address corresponding to this article submission to download your certificate.