Person Identification Using Landmarks and Deep Learning Techniques

Vamadevan, Arundev (2023) Person Identification Using Landmarks and Deep Learning Techniques. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (660kB) | Preview

Abstract

This research focuses on person identification through the detection of landmarks from facial images. This study uses VGG-16 model for facial landmark extraction and the Support Vector Machines (SVM) classifier for person identification. The models are trained with dataset VGGFace-2 which contains a large collection of different personalities with large variations in pose, ethnicity, age and illumination. Multi-Task cascaded Convolutional Neural Network (MTCNN) is used for face detection from the images. VGG-16 model is validated with a Root Mean Squared Error (RMSE) of 7.02 for landmark extraction and SVM identified persons from images with an accuracy of 70 %. The five facial key points selected are landmark coordinates of eyes, nose and distance between lip to lip. Histogram of Oriented Gradients (HOG) is used to calculate the feature descriptors which is used to train the SVM model. Images and facial landmarks are transformed using augmentation techniques. A comprehensive and detailed evaluation is conducted and the result shows a better performance in landmark extraction with deep learning models.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Horn, Christian UNSPECIFIED
Uncontrolled Keywords:	Landmark detection; VGG-16; SVM; MTCNN; VGGFace-2
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	23 May 2025 14:50
Last Modified:	23 May 2025 14:50
URI:	https://norma.ncirl.ie/id/eprint/7629

Actions (login required)

View Item