Efficient on-device saliency prediction via knowledge distillation and explainable AI (XAI) /
Files
(Published version)
Date
2024
Authors
Umer, Ayaz
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
thesis
Citation
Statement of Responsibility
Conference Name
Abstract
Humans can analyse complex visual scenes by selectively focusing their attention on specific regions. This selective attention mechanism of the human visual system, when mimicked by a computer through a simulation is known as saliency prediction or visual attention prediction. This mechanism plays an important role in many applications, including autonomous driving, advertisements, defence, medical, and human-computer interaction. This work proposed novel methods to facilitate the development of saliency prediction networks with reduced computational cost by utilising knowledge distillation. Furthermore, insight into the distillation method plays an important role in enhancing transparency and interpretability. Thus, by leveraging Explainable AI, insight is provided into how knowledge is transferred from the teacher to the student network under the knowledge distillation method.
School/Discipline
University of South Australia. UniSA STEM
UniSA STEM
UniSA STEM
Dissertation Note
Thesis (PhD(Computer and Information Science))--University of South Australia, 2024.
Provenance
Copyright 2024 Ayaz Umer
Description
1 ethesis (xvii, 121 pages) :
colour illustrations.
Includes bibliographical references (pages 104-117)
colour illustrations.
Includes bibliographical references (pages 104-117)
Access Status
506 0#$fstar $2Unrestricted online access