This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Adapting the models for human performanceAdapting the models for human performanceAvraham, Yeshurun & Lindenbaum , Journal of Vision 2008Avraham, Yeshurun & Lindenbaum , Journal of Vision 2008
► Study 2:Study 2:
Extending the models to account for spatial Extending the models to account for spatial organization effects.organization effects.
The FLNN modelThe FLNN model
feature extraction
ix
Farthest Labeled Nearest Neighbor
feature space
Avraham & Lindenbaum, IEEE-PAMI 2006
0° 30° 60° orientation
T = 0° D = 30°, 60°
T D D
The FLNN model – contThe FLNN model – cont..
1
2
3
4
Avraham & Lindenbaum, IEEE-PAMI 2006
Alternative parallel explanation: dynamic priority map
► Homogeneous distractors Homogeneous distractors
► Clustered distractors Clustered distractors –– maximum one from each maximum one from each clustercluster
Qualitative model behaviorQualitative model behavior
D
T
D
T
D
D
1
2
3
1
2
– pop-out behavior
Avraham & Lindenbaum, IEEE-PAMI 2006
Visual Search DifficultyVisual Search Difficulty►Search difficulty depends on two factors:Search difficulty depends on two factors:
Avraham, Yeshurun & Lindenbaum, Journal of Vision 2008
Study 1 SummaryStudy 1 Summary
► FLNN and COVER predict T-D and D-D similarity FLNN and COVER predict T-D and D-D similarity effects better than other prominent effects better than other prominent computational models.computational models.
► The models quantify grouping-by-similarity The models quantify grouping-by-similarity involved in visual search, by suggesting that the involved in visual search, by suggesting that the degree of within-group heterogeneity depends degree of within-group heterogeneity depends on the T-D similarity.on the T-D similarity.
► One possibility: a multi scale approachOne possibility: a multi scale approach((e.g., Itti et al. 1998, Rosenholtz et al. 2007e.g., Itti et al. 1998, Rosenholtz et al. 2007))
How to combine the measure over scales? How to combine the measure over scales?
max? weight and sum?max? weight and sum?Avraham, Yeshurun, Lindenbaum VSS 2011
► AdvantagesAdvantages: : same treatment for similarity and proximity same treatment for similarity and proximity understand and quantify the relative effect of eachunderstand and quantify the relative effect of each
► Questions to answer in this study:Questions to answer in this study: Will this enable our models to account for the Will this enable our models to account for the
spatial organization effect?spatial organization effect? What is the value of ?What is the value of ? Is stable or stimuli dependent?Is stable or stimuli dependent?
Avraham, Yeshurun, Lindenbaum VSS 2011
, feature spatial(1 )i jD d d
FLNN predictionsFLNN predictions
► Conclusions:Conclusions: Combination of feature distance and spatial distance is Combination of feature distance and spatial distance is
essential for predictionessential for prediction Limited possibilities: implies that the model is informativeLimited possibilities: implies that the model is informative Relates to previous findings regarding the combined effects Relates to previous findings regarding the combined effects
of proximity and similarity on perceptual groupingof proximity and similarity on perceptual grouping
((e.g., Kobovy & van den Berg 2008e.g., Kobovy & van den Berg 2008))Avraham, Yeshurun, Lindenbaum VSS 2011
0.35 prediction prediction withwith
predictive ability vs. predictive ability vs.
Maximum value of 2 to pass the 2 test
Preliminary: Is stable Preliminary: Is stable or stimuli dependent?or stimuli dependent?
► Experiment 2: Experiment 2: Manipulate the number of distractor typesManipulate the number of distractor types
► Experiment 3:Experiment 3:Manipulate the distractors varianceManipulate the distractors variance
Avraham, Yeshurun, Lindenbaum VSS 2011
Exp 2 (preliminary): Exp 2 (preliminary): manipulating the number of manipulating the number of
SummarySummary► A study of the effects of spatial organization on A study of the effects of spatial organization on
visual searchvisual search
► The FLNN model can predict effects of grouping The FLNN model can predict effects of grouping by similarity and grouping be proximityby similarity and grouping be proximity
► As it uses an explicit combination of feature As it uses an explicit combination of feature difference and spatial distance, it can help us difference and spatial distance, it can help us understand the relative effect of similarity and understand the relative effect of similarity and proximity on visual searchproximity on visual search