{"id":37751,"date":"2015-07-27T07:29:11","date_gmt":"2015-07-27T07:29:11","guid":{"rendered":"http:\/\/www.labri.fr\/perso\/barla\/blog\/?p=37751"},"modified":"2018-09-06T12:10:07","modified_gmt":"2018-09-06T12:10:07","slug":"exp-w-dicle","status":"publish","type":"post","link":"https:\/\/www.labri.fr\/perso\/barla\/blog\/?p=37751","title":{"rendered":"Exp w\/ Dicle"},"content":{"rendered":"<p id=\"top\" \/>&nbsp;<\/p>\n<h2>Experiments<\/h2>\n<p>Here is a list of four pipeline\/experiments (we talked about #1 to #3 yesterday, didn&#8217;t have time for #4).<br \/>\nI realized only now that two are about motion, two about shape, each with a sticky and a mirror material config.<\/p>\n<h3>#1 Rotation\/Translation ambiguity<\/h3>\n<p>Setup: bumpy plane seen up close, with stiky (also mirror?) material, translating horizontaly<br \/>\nVariations: increase the camera field of view and\/or surface bumpiness<br \/>\nQuestion: is motion perceived as a rotation or translation (2AFC) ?<br \/>\nHypothesis: high FoV &amp; bumpiness increase slant induced distortions -&gt; rotation around Y<\/p>\n<h3>#2 Rigid\/Non-rigid ambiguity<\/h3>\n<p>Setup: bumpy sphere seen entirely, with mirror material, rotating around Z in rocking motion<br \/>\nVariations: increase bumpiness &amp; anisotropy, tilt away from Z axis<br \/>\nQuestion: is motion perceived as rigid or non-rigid (2AFC) ? or not even 3D (3AFC) ?<br \/>\nHypothesis: stronger elongated bumps produce more stable regions -&gt; pooled to yield 3D percept.<\/p>\n<h3>#3 Global shape ambiguity<\/h3>\n<p>Setup: 3D ellipsoid w\/ horizontal bumps in the middle, sharp circular mask, sticky (also mirror?) materials, rotation around X<br \/>\nVariations: increase length of ellipsoid along horizontal axis, vary radius of sharp mask<br \/>\nQuestion: is shape behind aperture perceived as a sphere or cylinder (2AFC) ? or inbetween (3, 4 or 5 AFC) ?<br \/>\nHypothesis: when change of slant-induced distortions along X is strong enough -&gt; spherical shape.<\/p>\n<h3>#4 Local shape ambiguity<\/h3>\n<p>Setup: blobby\/bumpy object seen entirely, mirror material, rotations of both object AND lighting.<br \/>\nVariations: increase bumpiness and anisotropy, change angle between the two rotation axis.<br \/>\nQuestion: how much bumpy is the object ? (no idea what experimental method is best here)<br \/>\nHypothesis: stronger elongated bumps in direction of (object or lighting?) motion -&gt; more veridical shape percept.<\/p>\n<p>&nbsp;<\/p>\n<h2>Observations<\/h2>\n<h3>Sticky (non-homogeneous matte object in uniform lighting)<\/h3>\n<p>&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt; Rotation around X or Y &gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;<\/p>\n<p>* wMask \/ oneDir<br \/>\n&gt; Motion: translation at low bump levels, magnitude and slant (along motion) help perceive rotation<br \/>\n&gt; Shape: good overall, texture contrast helps perceive bumpiness<\/p>\n<p>* wMask \/ twoDir<br \/>\n&gt; Motion: maybe makes rotation more perceivable ?<br \/>\n&gt; Shape: no effect here<\/p>\n<p>* woMask \/ oneDir &amp; twoDir<br \/>\n&gt; Motion: always rotation, no effect of shape on perception<br \/>\n&gt; Shape: no ambiguity, some effect of texture contrast on bumpiness<\/p>\n<p>&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt; Rotation around Z &gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;<\/p>\n<p>* wMask \/ oneDir &amp; twoDir<br \/>\n&gt; Motion: simple 2D rotation<br \/>\n&gt; Shape: none, even with high bumpiness<\/p>\n<p>* woMask \/ oneDir &amp; twoDir<br \/>\n&gt; Motion: 3D rotation<br \/>\n&gt; Shape: same as with rotation around X or Y<\/p>\n<p>&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. Suggestions &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;<\/p>\n<p>&#8211; translation is perceived when compression\/stretching along motion direction is small.<br \/>\n&#8211; with isotropic bumps of low magnitude, distortions may be attributed\u00a0 to parallax<br \/>\n&#8211; with strong anisotropic bumps, especially along rotation dir, rotation gets most probable<br \/>\n&#8211; texture contrast probably helps the noticing of distortions by attracting attention<br \/>\n&#8211; rotation around Z has no distortion; it relies only on silhouettes to convey 3D shape.<br \/>\n&#8211; silhouettes (X&amp;Y) bring many more cues: (dis-)occlusions &amp; strong distortion gradients close to axis.<br \/>\n&#8211; this is why the addition of a hard mask make results look 2D entirely.<br \/>\n&#8211; no significant effect of reversibility because all motions are perceived rigid here (2D or 3D)&#8230;<br \/>\n&gt; seems that slant in the direction of motion controls a decisive compression\/stretching pdeudocue.<br \/>\n&gt; the pattern of slant-induced distortions varies across the object, abruptly when close to rotation axis.<br \/>\n&gt; This pattern not only conveys 3D motion cues (requiring to pool distortions across the image)<br \/>\n&gt; It also seems to convey 3D shape cues through the &#8220;signed&#8221; distorsions in the image<\/p>\n<p>&nbsp;<\/p>\n<h3>\u00a0Specular (homogeneous mirror object in noisy lighting<\/h3>\n<p>&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt; Rotation around X or Y &gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;<\/p>\n<p>* wMask \/ oneDir<br \/>\n&gt; Motion: 2D translation+non-rigid at all bump levels except 5 maybe, high curvature magnitude &amp; anisotropy necessary for rotation<br \/>\n&gt; Shape: weird flat pattern at low bump levels, motion perpendicular to long bumps (stability) necessary for 3D perception<\/p>\n<p>* wMask \/ twoDir<br \/>\n&gt; Motion: Not much amelioration in the perception of rotations<br \/>\n&gt; Shape: 3D percept increase as stable parts are more easily picked up<\/p>\n<p>* woMask \/ oneDir &amp; twoDir<br \/>\n&gt; Motion: always rotation, no effect of shape on perception<br \/>\n&gt; Shape: bumps of small magnitude hard to follow (changing contrast), better w\/ twoDir<\/p>\n<p>&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt;&lt; Rotation around Z &gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;&gt;<\/p>\n<p>* wMask \/ oneDir<br \/>\n&gt; Motion: simple 2D rotation<br \/>\n&gt; Shape: none, even with high bumpiness<\/p>\n<p>* wMask \/ twoDir<br \/>\n&gt; Motion: 2D \/ 3D rotation, not clear.<br \/>\n&gt; Shape: weak 3D percept for high bumps<\/p>\n<p>* woMask \/ oneDir<br \/>\n&gt; Motion: 3D rotation at high bumpiness levels, but isotropic bumps seem to move around at low levels.<br \/>\n&gt; Shape: low bumps seem to appear or disappear depending on contrast, anisotropy helps resolve that.<\/p>\n<p>* woMask \/ twoDir<br \/>\n&gt; Motion: full rigid rotation only at bump level 5, otherwise multiple rotations at different rates !<br \/>\n&gt; Shape: minimize the disappearance of low bumps, all bumps more visible here!<\/p>\n<p>&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. Suggestions &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.<\/p>\n<p>&#8211; stability at isotropic bumps is not enough (2D translation), requires strong anisotropy to get a full line feature rotating in 3D<br \/>\n&#8211; hence only with strong bumps do we get a 3D percept, so that there is more stability than sliding in the motion pattern<br \/>\n&#8211; reversibility helps capture the coherent motion of stability regions, hence suggesting a more 3D-like percept.<br \/>\n&#8211; also happens without mask, since patterns of low contrast may force low isotropic bumps to go unnoticed.<br \/>\n&#8211; silhouettes have a dramatic effect on motion perception, which might be due to differential motion of coherent stable points<br \/>\n&#8211; but also dramatic on shape perception, even though the motion patterns appear quite complex along the boundary&#8230;<br \/>\n&#8211; silhouette cliping has a brutal effect on rotations around Z, killing the perception of 3D shape and 3D motion altogether<br \/>\n&#8211; here reversibility permits to better perceive bumps, but at low levels they remain too isolated to give a clear 3D motion.<br \/>\n&#8211; surprisingly, enabling silhouettes gives a non-rigid combination of rotations, all the more visible with reversible motion !<br \/>\n&gt; seems that curvature anisotropy w.r.t. the direction of motion controls a decisive stability\/sliding pseudo-cue.<br \/>\n&gt; the pattern of curvature-induced distortions varies across the object, with more stable regions close to the rotation axis.<br \/>\n&gt; this pattern not only conveys 3D shape cues (up to a convex\/concave ambiguity when silhouettes are not to be seen).<br \/>\n&gt; but also 3D motion cues, requiring to pool stable regions across the image (which is prone to fail in non-generic motion).<\/p>\n","protected":false},"excerpt":{"rendered":"<p>&nbsp; Experiments Here is a list of four pipeline\/experiments (we talked about #1 to #3 yesterday, didn&#8217;t have time for #4). I realized only now that two are about motion, two about shape, each with a sticky and a mirror material config. #1 Rotation\/Translation ambiguity Setup: bumpy plane seen up close, with stiky (also mirror?) &#8230; <a title=\"Exp w\/ Dicle\" class=\"read-more\" href=\"https:\/\/www.labri.fr\/perso\/barla\/blog\/?p=37751\" aria-label=\"Read more about Exp w\/ Dicle\">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":[641],"tags":[],"class_list":["post-37751","post","type-post","status-publish","format-standard","hentry","category-tech"],"_links":{"self":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/37751","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=37751"}],"version-history":[{"count":3,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/37751\/revisions"}],"predecessor-version":[{"id":39814,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=\/wp\/v2\/posts\/37751\/revisions\/39814"}],"wp:attachment":[{"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=37751"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=37751"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.labri.fr\/perso\/barla\/blog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=37751"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}