A computational model of task-dependent influences on eye position

A computational model of task-dependent influences on eye position. Peters, R. J. & Itti, L. In Proc. Vision Science Society Annual Meeting (VSS06), May, 2006.
abstract bibtex

Computational models of bottom-up attention can perform significantly above chance at predicting eye positions of observers passively viewing static or dynamic images. Nevertheless, much of eye movement behavior (50 percent or more) is unexplained by purely bottom-up models, and is typically attributed to top-down, inter-observer, task-dependent, or random effects. Other studies have qualitatively described such high-level effects in naturalistic interactive visual tasks (e.g., while driving, how often do people fixate other cars, or the road, or road signs); yet the underlying neurocomputational mechanisms remain unknown. Here, we introduce a simple computational model of task-related eye position influences in interactive tasks with dynamic stimuli. This model extracts from each frame a low-dimensional feature signature ("gist"), compares that with a database of eye position training frames, and produces an eye position prediction map. Finally, we combine the task-related and bottom-up maps, and compare the combined maps with observers' actual eye positions across 216,000 frames from 24 five-minute videogame-playing sessions. For analysis, each map was rescaled to have zero mean and unit standard deviation; the average predicted value at human eye position locations was 0.61 +/- 0.1 in the purely bottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model gives an average value of 0). Thus, this straightforward model of task-dependent effects offers some of the strongest purely computational general-purpose eye movement predictions to date, going significantly beyond what is explained by purely bottom-up effects; yet it relies only on simple visual features, without requiring any high-level semantic scene description.

@inproceedings{ Peters_Itti06vss,
  author = {R. J. Peters and L. Itti},
  title = {A computational model of task-dependent influences on eye position},
  abstract = {Computational models of bottom-up attention can perform
significantly above chance at predicting eye positions of observers
passively viewing static or dynamic images. Nevertheless, much of eye
movement behavior (50 percent or more) is unexplained by purely
bottom-up models, and is typically attributed to top-down,
inter-observer, task-dependent, or random effects. Other studies have
qualitatively described such high-level effects in naturalistic
interactive visual tasks (e.g., while driving, how often do people
fixate other cars, or the road, or road signs); yet the underlying
neurocomputational mechanisms remain unknown. Here, we introduce a
simple computational model of task-related eye position influences in
interactive tasks with dynamic stimuli. This model extracts from each
frame a low-dimensional feature signature ("gist"), compares that
with a database of eye position training frames, and produces an eye
position prediction map. Finally, we combine the task-related and
bottom-up maps, and compare the combined maps with observers' actual
eye positions across 216,000 frames from 24 five-minute
videogame-playing sessions. For analysis, each map was rescaled to
have zero mean and unit standard deviation; the average predicted
value at human eye position locations was 0.61 +/- 0.1 in the purely
bottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model
gives an average value of 0). Thus, this straightforward model of
task-dependent effects offers some of the strongest purely
computational general-purpose eye movement predictions to date, going
significantly beyond what is explained by purely bottom-up effects;
yet it relies only on simple visual features, without requiring any
high-level semantic scene description.},
  booktitle = {Proc. Vision Science Society Annual Meeting (VSS06)},
  year = {2006},
  month = {May},
  type = {mod;bu;td;eye},
  review = {abs/conf}
}

Downloads: 0

{"_id":{"_str":"5298a1a19eb585cc260008ec"},"__v":0,"authorIDs":[],"author_short":["Peters, R.<nbsp>J.","Itti, L."],"bibbaseid":"peters-itti-acomputationalmodeloftaskdependentinfluencesoneyeposition-2006","bibdata":{"html":"<div class=\"bibbase_paper\"> \n\n\n<span class=\"bibbase_paper_titleauthoryear\">\n\t<span class=\"bibbase_paper_title\"><a name=\"Peters_Itti06vss\"> </a>A computational model of task-dependent influences on eye position.</span>\n\t<span class=\"bibbase_paper_author\">\nPeters, R. J.; and Itti, L.</span>\n\t\n</span>\n\n\n\nIn\n<i>Proc. Vision Science Society Annual Meeting (VSS06)</i>, May 2006.\n\n\n\n\n\n<br class=\"bibbase_paper_content\"/>\n\n<span class=\"bibbase_paper_content\">\n \n \n \n <a href=\"javascript:showBib('Peters_Itti06vss')\"\n class=\"bibbase link\">\n \n\t\n\t\n\t\n BibTeX\n <i class=\"fa fa-caret-down\"></i></a>\n \n \n  \n <a class=\"bibbase_abstract_link bibbase link\"\n href=\"javascript:showAbstract('Peters_Itti06vss')\">\n Abstract\n <i class=\"fa fa-caret-down\"></i></a>\n \n \n \n\n \n \n \n</span>\n\n<div class=\"well well-small bibbase\" id=\"bib_Peters_Itti06vss\"\n style=\"display:none\">\n <pre>@inproceedings{ Peters_Itti06vss,\n author = {R. J. Peters and L. Itti},\n title = {A computational model of task-dependent influences on eye position},\n abstract = {Computational models of bottom-up attention can perform\nsignificantly above chance at predicting eye positions of observers\npassively viewing static or dynamic images. Nevertheless, much of eye\nmovement behavior (50 percent or more) is unexplained by purely\nbottom-up models, and is typically attributed to top-down,\ninter-observer, task-dependent, or random effects. Other studies have\nqualitatively described such high-level effects in naturalistic\ninteractive visual tasks (e.g., while driving, how often do people\nfixate other cars, or the road, or road signs); yet the underlying\nneurocomputational mechanisms remain unknown. Here, we introduce a\nsimple computational model of task-related eye position influences in\ninteractive tasks with dynamic stimuli. This model extracts from each\nframe a low-dimensional feature signature (\"gist\"), compares that\nwith a database of eye position training frames, and produces an eye\nposition prediction map. Finally, we combine the task-related and\nbottom-up maps, and compare the combined maps with observers' actual\neye positions across 216,000 frames from 24 five-minute\nvideogame-playing sessions. For analysis, each map was rescaled to\nhave zero mean and unit standard deviation; the average predicted\nvalue at human eye position locations was 0.61 +/- 0.1 in the purely\nbottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model\ngives an average value of 0). Thus, this straightforward model of\ntask-dependent effects offers some of the strongest purely\ncomputational general-purpose eye movement predictions to date, going\nsignificantly beyond what is explained by purely bottom-up effects;\nyet it relies only on simple visual features, without requiring any\nhigh-level semantic scene description.},\n booktitle = {Proc. Vision Science Society Annual Meeting (VSS06)},\n year = {2006},\n month = {May},\n type = {mod;bu;td;eye},\n review = {abs/conf}\n}</pre>\n</div>\n\n\n<div class=\"well well-small bibbase\" id=\"abstract_Peters_Itti06vss\"\n style=\"display:none\">\n Computational models of bottom-up attention can perform significantly above chance at predicting eye positions of observers passively viewing static or dynamic images. Nevertheless, much of eye movement behavior (50 percent or more) is unexplained by purely bottom-up models, and is typically attributed to top-down, inter-observer, task-dependent, or random effects. Other studies have qualitatively described such high-level effects in naturalistic interactive visual tasks (e.g., while driving, how often do people fixate other cars, or the road, or road signs); yet the underlying neurocomputational mechanisms remain unknown. Here, we introduce a simple computational model of task-related eye position influences in interactive tasks with dynamic stimuli. This model extracts from each frame a low-dimensional feature signature (\"gist\"), compares that with a database of eye position training frames, and produces an eye position prediction map. Finally, we combine the task-related and bottom-up maps, and compare the combined maps with observers' actual eye positions across 216,000 frames from 24 five-minute videogame-playing sessions. For analysis, each map was rescaled to have zero mean and unit standard deviation; the average predicted value at human eye position locations was 0.61 +/- 0.1 in the purely bottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model gives an average value of 0). Thus, this straightforward model of task-dependent effects offers some of the strongest purely computational general-purpose eye movement predictions to date, going significantly beyond what is explained by purely bottom-up effects; yet it relies only on simple visual features, without requiring any high-level semantic scene description.\n</div>\n\n\n</div>\n","downloads":0,"bibbaseid":"peters-itti-acomputationalmodeloftaskdependentinfluencesoneyeposition-2006","role":"author","year":"2006","type":"mod;bu;td;eye","title":"A computational model of task-dependent influences on eye position","review":"abs/conf","month":"May","key":"Peters_Itti06vss","id":"Peters_Itti06vss","booktitle":"Proc. Vision Science Society Annual Meeting (VSS06)","bibtype":"inproceedings","bibtex":"@inproceedings{ Peters_Itti06vss,\n author = {R. J. Peters and L. Itti},\n title = {A computational model of task-dependent influences on eye position},\n abstract = {Computational models of bottom-up attention can perform\nsignificantly above chance at predicting eye positions of observers\npassively viewing static or dynamic images. Nevertheless, much of eye\nmovement behavior (50 percent or more) is unexplained by purely\nbottom-up models, and is typically attributed to top-down,\ninter-observer, task-dependent, or random effects. Other studies have\nqualitatively described such high-level effects in naturalistic\ninteractive visual tasks (e.g., while driving, how often do people\nfixate other cars, or the road, or road signs); yet the underlying\nneurocomputational mechanisms remain unknown. Here, we introduce a\nsimple computational model of task-related eye position influences in\ninteractive tasks with dynamic stimuli. This model extracts from each\nframe a low-dimensional feature signature (\"gist\"), compares that\nwith a database of eye position training frames, and produces an eye\nposition prediction map. Finally, we combine the task-related and\nbottom-up maps, and compare the combined maps with observers' actual\neye positions across 216,000 frames from 24 five-minute\nvideogame-playing sessions. For analysis, each map was rescaled to\nhave zero mean and unit standard deviation; the average predicted\nvalue at human eye position locations was 0.61 +/- 0.1 in the purely\nbottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model\ngives an average value of 0). Thus, this straightforward model of\ntask-dependent effects offers some of the strongest purely\ncomputational general-purpose eye movement predictions to date, going\nsignificantly beyond what is explained by purely bottom-up effects;\nyet it relies only on simple visual features, without requiring any\nhigh-level semantic scene description.},\n booktitle = {Proc. Vision Science Society Annual Meeting (VSS06)},\n year = {2006},\n month = {May},\n type = {mod;bu;td;eye},\n review = {abs/conf}\n}","author_short":["Peters, R.<nbsp>J.","Itti, L."],"author":["Peters, R. J.","Itti, L."],"abstract":"Computational models of bottom-up attention can perform significantly above chance at predicting eye positions of observers passively viewing static or dynamic images. Nevertheless, much of eye movement behavior (50 percent or more) is unexplained by purely bottom-up models, and is typically attributed to top-down, inter-observer, task-dependent, or random effects. Other studies have qualitatively described such high-level effects in naturalistic interactive visual tasks (e.g., while driving, how often do people fixate other cars, or the road, or road signs); yet the underlying neurocomputational mechanisms remain unknown. Here, we introduce a simple computational model of task-related eye position influences in interactive tasks with dynamic stimuli. This model extracts from each frame a low-dimensional feature signature (\"gist\"), compares that with a database of eye position training frames, and produces an eye position prediction map. Finally, we combine the task-related and bottom-up maps, and compare the combined maps with observers' actual eye positions across 216,000 frames from 24 five-minute videogame-playing sessions. For analysis, each map was rescaled to have zero mean and unit standard deviation; the average predicted value at human eye position locations was 0.61 +/- 0.1 in the purely bottom-up maps, and 2.42 +/- 0.07 in the combined maps (a random model gives an average value of 0). Thus, this straightforward model of task-dependent effects offers some of the strongest purely computational general-purpose eye movement predictions to date, going significantly beyond what is explained by purely bottom-up effects; yet it relies only on simple visual features, without requiring any high-level semantic scene description."},"bibtype":"inproceedings","biburl":"http://ilab.usc.edu/publications/src/ilab.bib","downloads":0,"search_terms":["computational","model","task","dependent","influences","eye","position","peters","itti"],"title":"A computational model of task-dependent influences on eye position","year":2006,"dataSources":["wedBDxEpNXNCLZ2sZ"]}