{"id":771,"date":"2010-06-18T12:25:38","date_gmt":"2010-06-18T12:25:38","guid":{"rendered":"http:\/\/www.labri.fr\/perso\/barla\/blog\/?p=771"},"modified":"2011-06-07T09:50:59","modified_gmt":"2011-06-07T09:50:59","slug":"neurocomp-2008-workshop-on-computational-methods-for-biological-vision","status":"publish","type":"post","link":"https:\/\/www.labri.fr\/perso\/barla\/blog\/?p=771","title":{"rendered":"Neurocomp 2008"},"content":{"rendered":"<p id=\"top\" \/>\n<h2>Extraction of computational principles from the biological study of sensory cortical dynamics (Yves Fr\u00e9gnac)<\/h2>\n<ul>\n<li>L+NL model for representing simple and complex cells<\/li>\n<li>Linearities model single cells in some conditions<\/li>\n<li>Non-linearities account for scale and orientation selectivity<\/li>\n<li>But dynamic non-linearities occur, especially for natural images<\/li>\n<li>Dynamic behaviors seem to depend on context<\/li>\n<li>Need a richer model (excitatory and suppressive, on and off, L+NL networks)<\/li>\n<li>Noise also seems to play an important role<\/li>\n<li>Up and down states impose conditions on cell behaviors (do they create propagation waves on neighboring cells?)<\/li>\n<li>Cells response depend on input spectrum<\/li>\n<li>Temporal arrangements between excitation and inhibition condition cell response (inhibition follows excitation slightly)<\/li>\n<li>Try to find the details of the network dynamics as they have been trained through learning or epigenesis, and provide models for this.<\/li>\n<\/ul>\n<h2>Neural fields and the modelling of primary visual cortex (Olivier Faugeras)<\/h2>\n<ul>\n<li>Cortical columns are ill-defined (different types)<\/li>\n<li>Neural mass models represent average neuronal activity<\/li>\n<li>Continuum or cortical columns modeled by a neural mass recursive equation<\/li>\n<li>The contribution of neighboring cells is modelled via a L+NL system, but the L part is partially unkown<\/li>\n<li>Adding orientation features to the model, they make the distinction between the local and distant lateral connections (Gestalt)<\/li>\n<li>Stationary solutions give an explanation to junctions (probably thanks to lateral connections)<\/li>\n<li>Spatial frequency is modelled by the two pinwheels of each hyper-column (like a frequency range)<\/li>\n<\/ul>\n<h2>Active perception and visual attention (Nicolas Rougier)<\/h2>\n<ul>\n<li>Difference between parallel and sequential search.<\/li>\n<li>Computational model of focus based on discretized neural fields equations solved in parallel<\/li>\n<li>Introduce anticipation and memory in the system.<\/li>\n<\/ul>\n<h2>Temporal dynamics of 2D motion integration (G. Masson)<\/h2>\n<ul>\n<li>Avoid mixing ground and figure motion<\/li>\n<li>Deal with aperture problem<\/li>\n<li>Make 2D rigidity hypothesis<\/li>\n<li>Rely on eye pursuit to progressively resolve ambiguity<\/li>\n<li>Use a bayesian approach to model the pursuit<\/li>\n<li>Introduce extra-retinal (motor?) information<\/li>\n<li>Show that there is no anticipatory motion based on cues to solve the aperture problem<\/li>\n<li>Conclude that MST is a gear between image and object motion, which provides reactivity.<\/li>\n<\/ul>\n<h2>Biological stereopsis from the computer vision viewpoint (Miles Hansard &amp; Radu Horaud)<\/h2>\n<ul>\n<li>Many disparity measurements, but some are false (depend on binocular gaze) and many are not trivial<\/li>\n<li>Computer vision rather uses epipolar geometry<\/li>\n<li>Interesting for biology because it helps understand stimulii<\/li>\n<li>Of course, eye kinematics are limited in Vision<\/li>\n<li>Describe epipolar geometry of stereo fixation<\/li>\n<li>Use a preliminary problem approach to predict dispaity neurons connections<\/li>\n<\/ul>\n<h2>Stereopsis from the Neurophysiologist viewpoint (Yves Trotter)<\/h2>\n<ul>\n<li>Combination and processing of stereopsis cues<\/li>\n<li>Correction of depth occurs to compensate an inverse distance law<\/li>\n<li>Various cells are tund to different disparities (excitatory and inhibitory)<\/li>\n<li>Use of extra-retinal information on eye gaze modifies the gain so that perception of depth is coherent<\/li>\n<li>Vertical disparities use same cells as horizontal ones<\/li>\n<li>Natural and artificial stimuli seem to be processed differently<\/li>\n<\/ul>\n<h2>Retina, neuromorphic circuits and fundamental properties (Jeanny Herault)<\/h2>\n<ul>\n<li>Retina performs a preprocessing of optical information<\/li>\n<li>Horizontal cells help perform local contrast enhancement<\/li>\n<li>Bipolar cells accumulate receptor output with increasing excentricity<\/li>\n<li>Amacrine cells perform a high pass temporal processing<\/li>\n<li>Inter-plexiform cells help adaptation by controlling horizontal cells<\/li>\n<li>Ganglion cells send spikes to different channels of LGN<\/li>\n<li>Output of bipolar cells perform high pass spatio-temporal filtering ; compensates the 1\/f spectrum of natural images<\/li>\n<li>H-cells compute the mean, ganglion cells the local centering, and magno local variance (in time also)<\/li>\n<li>Separation of chrominance and luminance is done at the cortical level<\/li>\n<li>Random sampling of colors help reduce aliasing<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Extraction of computational principles from the biological study of sensory cortical dynamics (Yves Fr\u00e9gnac) L+NL model for representing simple and complex cells Linearities model single cells in some conditions Non-linearities account for scale and orientation selectivity But dynamic non-linearities occur, especially for natural images Dynamic behaviors seem to depend on context Need a richer model &#8230; <a title=\"Neurocomp 2008\" class=\"read-more\" href=\"https:\/\/www.labri.fr\/perso\/barla\/blog\/?p=771\" aria-label=\"Read more about Neurocomp 2008\">Read more<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[81],"tags":[],"class_list":["post-771","post","type-post","status-publish","format-standard","hentry","category-talks"],"_links":{"self":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/771","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=771"}],"version-history":[{"count":8,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/771\/revisions"}],"predecessor-version":[{"id":14281,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/771\/revisions\/14281"}],"wp:attachment":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=771"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=771"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=771"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}