Dan Kersten Computational Vision Lab Psychology Department, U. Minnesota SUnS kersten.org

Transcription

1 How big is it? Dan Kersten Computational Vision Lab Psychology Department, U. Minnesota SUnS 2009 kersten.org NIH R01 EY NIH P , P30 NS and the MIND Institute

2 Huseyin Boyaci Bilkent U, Ankara Fang Fang, Beijing U Scott Murray U Wash Murray, S. O., Boyaci, H., & Kersten, D. (2006). The representation of perceived angular size in human primary visual cortex. Nat Neurosci, 9(3), Fang, F., Boyaci, H., Kersten, D., & Murray, S. O. (2008). Attention-dependent representation of a size illusion in human V1. Curr Biol, 18(21),

3 Bozeman, Montana vision experts were asked...

4 ...where would the top of your head be if you stood by that tree trunk? This was their guess

5 Reality ~ x 2 bigger

6 Reality Why? Clue: Statistical properties of scree fields

7 The size-depth ambiguity For a given retinal size, one could be looking at a big far object, or a near big object We can resolve size ambiguities using Prior expectations on object categories learned from experience 3D contextual cues for distance and size comparisons

8 ...not a newly recognized problem Moon illusion Aristotle, da Vinci, Kepler, Berkeley,... Fig. 1. Regardless of its elevation, the distance between an observer (at the center of the horizontal line) and the moon remains constant (unfilled circles). However, a moon perceived as growing closer as its elevation increases (filled circles), must appear as growing smaller. Al-Hazan s apparent distance theory, 11th century Ponzo illusion, 1913

9 Size is a problem of object: inference 3D scene context: size, category depth Generative view image size contextual cues image intensities

10 Size is a problem of object: inference 3D scene context: size, category depth Scene understanding problem Segmentation problem image size cues contextual cues image intensities

11 Cortical Areas? LOC? object: 3D scene context: PPA? size, category depth V1? image size cues contextual cues

12 Primary Visual Cortex: V1 Models begin with two basic assumptions: Local: small hypercolumns consisting of banks of neurons tuned for orientation, etc... that analyze points of images entering the eye Global: hypercolumns arranged retinotopically

13 Eccentricity--the radial direction Huk, A. C. (2008) Visual Neuroscience: Retinotopy meets Percept-otopy, Current Biology, 18, 21, R

14 3D influences on spatial extent of fmri BOLD? Measured activity shift in V1 in response to the edge of the checkered balls: Relative to 2D retinotopically mapped V1: Increase in apparent size led to spatial shift of peak activity away from foveal representation while a decrease led to a shift towards the foveal representation Murray, S. O., Boyaci, H., & Kersten, D. (2006). The representation of perceived angular size in human primary visual cortex. Nat Neurosci, 9(3),

15 Attention Use rings rather than disks to obtain a more direct measure of spatial shift -- cortical line-response Role of attention Fang, F., Boyaci, H., Kersten, D., & Murray, S. O. (2008). Attention-dependent representation of a size illusion in human V1. Curr Biol, 18(21),

16 Source of the modulation? Sources of depth information in a scene 3D context: linear perspective, texture, together with occlusion/shadows, relative size between visually similar objects Requires processes that compare and integrate over a large area of the visual field Integration relies on larger receptive fields at higher cortical areas of the visual system

17 Source of the modulation? Depth-related changes in V1 activity may depend on feedback from activity in higher visual areas Higher visual areas are strongly influenced by attention Can we silence attention? Two attention conditions attend-to-ring attend-to-fixation

18 Attention Attend to the center or attend to the ring

19 A Front fixation Static front ring Static back ring Front fixation Flickering front ring Static back ring Back fixation Static front ring Static back ring Back fixation Static front ring Flickering back ring 20 s 10 s 20 s 10 s 5 localizer rings

20 Shift to perceived size change Back Front

21 V1 line-response functions Back Front

22 Psychophysics Far ring at back appears about 15% bigger

23 Individual differences..suggestive, but need more subjects

24 Conclusions Attention to object/scene modulates spatial extent of BOLD activity in V1 Challenges What are the spatial functions of V1? Role for online computation of geometric relationships? Mechanisms underlying the interplay of retinotopic, object, and depth context information

25 How big is it? It s the same fish