Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/154983
Title: COMPUTATIONAL MODELS OF BOTTOM-UP AND TOP-DOWN VISUAL ATTENTION
Authors: ZHANG MENGMI
ORCID iD:   orcid.org/0000-0002-2694-7097
Keywords: visual attention, saliency, biologically-inspired vision, bottom-up attention, top-down attention, eye fixation
Issue Date: 17-Jan-2019
Citation: ZHANG MENGMI (2019-01-17). COMPUTATIONAL MODELS OF BOTTOM-UP AND TOP-DOWN VISUAL ATTENTION. ScholarBank@NUS Repository.
Abstract: Humans have the remarkable ability of prioritizing the sequence of eye-fixations to survey a complex environment. This ability enables us to react rapidly to environmental changes, and maximize the amount of useful information obtained from visual inputs, in spite of the limited high-acuity processing capability and memory capacity in the brain. This thesis describes several biologically-inspired computational implementations of integrated bottom-up and top-down visual attention that learn to exploit the temporal dynamics across fixations via supervised training, and modulate the visual processing pathway in a top-down fashion via zero-shot learning. These models not only contribute to the development of artificial intelligence in terms of state-of-the-art prediction of fixation locations and anticipation in images, as well as egocentric and third-person videos, but also provide insights into the mechanisms of human visual attention. The latter was demonstrated by closely approximating the behavior of human eye movements in a series of psychophysics experiments.
URI: https://scholarbank.nus.edu.sg/handle/10635/154983
Appears in Collections:Ph.D Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
ZhangMM_thesis.pdf24.34 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

47
checked on Sep 18, 2020

Download(s)

13
checked on Sep 18, 2020

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.