BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Cortical mechanisms underlying integration of local visual cues to
  form global representations - Dr Wei Wang
DTSTART:20181015T120000Z
DTEND:20181015T130000Z
UID:TALK110197@talks.cam.ac.uk
CONTACT:John Mollon
DESCRIPTION:Human and non‐human primates effortlessly see both global an
 d local features of objects in great detail. However\, how the cortex inte
 grates local visual cues to form global representations along visual hiera
 rchies remains mysterious\, particularly considering a long-standing parad
 ox in vision as neurally encoded complexity increases along the visual hie
 rarchy\, the known acuity or resolving power dramatically decreases. Putti
 ng it simply\, how do we simultaneously recognize the face of our child\, 
 while still resolving the individual hairs of her or his eyelashes? Many m
 odels of visual processing follow the idea that low-level resolution and p
 osition information is discarded to yield high-level representations (incl
 uding cutting edge deep learning models). These are themes that are fundam
 ental to conceiving how the brain does sensory transformation\, and curren
 t ideas require hypothetical complicated recurrent loops to connect global
  and local information across the visual hierarchy. \n\nCombining large-sc
 ale brain imaging to record the transformation of information across multi
 ple visual areas simultaneously with electrophysiological single-unit reco
 rdings (V1\, V2\, and V4\, MT/MST)\, we demonstrated a bottom-up cascade o
 f cortical visual processing mechanisms underlying integration of local in
 formation to form global representations in the primate ventral and dorsal
  visual streams. Our local-global stimuli are tyre-type text bars\, illuso
 ry contours\, and the Pinna-Brelstaff concentric rings. Recently\, we reve
 al an unexpected neural clustering preserving visual acuity from V1 into V
 4\, enabling a detailed spatiotemporal separation of local and global feat
 ures along the object-processing hierarchy\, suggesting that higher acuiti
 es are retained to later stages where more detailed cognitive behaviour oc
 curs. The study reinforces the point that neurons in V4 (and most likely a
 lso in infero-temporal cortex) do not necessarily need to have only low vi
 sual acuity\, which may begin to resolve the long-standing paradox concern
 ing fine visual discrimination. Thus\, our research will prompt further st
 udies to probe how preservation of low-level information is useful for hig
 her-level vision and provide new ideas to inspire the next generation of d
 eep neural network architectures. \n
LOCATION:Kenneth Craik Room\, Craik-Marshall Building\, Downing Site
END:VEVENT
END:VCALENDAR