NORMA eResearch @NCI Library

OpenPose based Gait Recognition using Triplet Loss Architecture

Shaik, Shahil (2020) OpenPose based Gait Recognition using Triplet Loss Architecture. Masters thesis, Dublin, National College of Ireland.

[img]
Preview
PDF (Master of Science)
Download (2MB) | Preview
[img]
Preview
PDF (Configuration manual)
Download (1MB) | Preview

Abstract

Behavioural biometrics have certain advantages over physiological biometrics. These biometrics does not require cooperative subject or proximity of individual. Gait recognition is form of behavioural biometrics where individual is identified based on their walking pattern. Most of the gait recognition models are based on temporal templates of human silhouettes such as gait energy image. In this project gait recognition using the OpenPose key point coordinates is proposed. OpenPose algorithm estimates the coordinates of various key points of individual in an image or video. Two approaches for feature extraction is applied on CASIA-B dataset. Manual features such as length of limbs and angle between limbs of individual are extracted using OpenPose key points coordinates. A 64-dimensional embedding vector is created for every video using deep learning triplet loss algorithm. Knn algorithm is trained on manual features and triplet loss features separately. An accuracy of 65 percent is achieved using the manual feature method and 71 percent using the triplet loss architecture.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science

Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Dan English
Date Deposited: 11 Jun 2020 13:49
Last Modified: 11 Jun 2020 13:49
URI: http://trap.ncirl.ie/id/eprint/4279

Actions (login required)

View Item View Item