BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260512T112204EDT-69894U913p@132.216.98.100
DTSTAMP:20260512T152204Z
DESCRIPTION:Abstract\n\nHuman visual attention has long been a topic of int
 erest for researchers and scientists. Visual attention is composed of fixa
 tions\, where the eyes remain stable to process the visual information\, a
 nd rapid eye movements between these fixations\, known as saccades. Fixati
 on points often carry a lot of information\, including key events within a
  scene and can offer insights into a viewer’s personality traits. Conseque
 ntly\, predicting the visual attention\, referred to as saliency predictio
 n\, has been a longstanding and significant research problem.\n\nIn the ar
 ea of saliency prediction\, most existing methods focus on universal salie
 ncy prediction\, the prediction of attention for an average viewer. These 
 methods fail to catch the inter-individual variability in attention. To ad
 dress this\, some methods have been proposed for personalized saliency pre
 diction\, which predict saliency for individuals by considering their feat
 ures. While these methods account for individual differences\, they face l
 imitations due to challenges in large-scale data collection\, noisy data\,
  and privacy concerns.\n\nTo address the issues associated with universal 
 and personalized saliency prediction\, this thesis presents methods for sa
 liency prediction in groups\, referred to as group saliency prediction. We
  propose grouping viewers based on similarities in demographics\, interest
 s\, visual attention\, and other available data. Based on these identified
  groups\, we design architectures for predicting saliency specific to each
  viewer group.\n\nOur first method is an image saliency prediction techniq
 ue called Clustered Saliency Prediction. This method groups viewers into c
 lusters based on their personal features and known saliency maps\, using s
 elected importance weights for personal feature factors. Building on these
  clusters\, we introduce the Multi-Domain Saliency Translation (MDST) mode
 l\, an image saliency prediction framework based on Generative Adversarial
  Networks (GANs)\, conditioned on cluster labels. The MDST model generates
  saliency maps tailored to each identified group of viewers. We evaluate o
 ur approach on a public dataset of personalized saliency maps and show tha
 t our method outperforms state-of-the-art universal saliency prediction mo
 dels. We also demonstrate the effectiveness of our clustering method by co
 mparing results using our clusters with those from baseline methods. Final
 ly\, we propose an approach to assign new individuals to their most approp
 riate cluster and show its applicability through a series of experiments.
 \n\nWe additionally introduce a novel set of generative neural networks de
 signed for saliency prediction tailored to viewer groups. These models are
  built on a generative framework that leverages style-transfer techniques 
 to transform universal saliency maps into group-specific predictions. We e
 valuate their performance on personalized saliency map datasets and invest
 igate the impact of data augmentation strategies. Additionally\, we analyz
 e the strengths and limitations of each model and conduct ablation studies
  to further justify our design decisions.\n\nLastly\, we apply our group s
 aliency prediction methods to a new egocentric video and eye-tracking data
 set that we acquired in a convenience store. This dataset includes 108 fir
 st-person videos of 36 shoppers searching for three products: orange juice
 \, KitKat chocolate bars\, and canned tuna\, along with eye fixation data 
 for each video frame. It also includes demographic information about each 
 participant in the form of an 11-question survey. Using survey responses\,
  our clustering method identified two distinct viewer groups. We trained o
 ur group saliency prediction models on the fixation data from the store vi
 deos. The results show improved saliency prediction performance on this re
 al-world dataset compared to leading universal models.\n
DTSTART:20250827T180000Z
DTEND:20250827T200000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD defence of Rezvan Sherkati – Saliency Prediction for Groups Usi
 ng Generative Models
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-rezvan-sherkati-sa
 liency-prediction-groups-using-generative-models-366388
END:VEVENT
END:VCALENDAR