This preview shows page 1. Sign up to view the full content.
Unformatted text preview: on-face images with varying amounts of noise were used in a face
tegorization task. b, Timing of events in each categorization trial. One of
e four possible microstimulation conditions shown was applied randomly
ehran, 19395, Iran. 2Research Center for Brain and Cognitive Sciences,
385, Iran. Afraz, Kiani, Esteky The visual system has to solve a hard problem: object recognition invariant to change Position Size Pose/Background We do not know how this is solved (we cannot build a machine that does it)
We have some hypotheses about how it might work The ventral visual pathway: selectivity and invariance
Macaque V4 V1 pIT
retina Selectivity Invariance Receptive ﬁelds
get bigger retina V1 V4 IT Position separabil P “Invariant” Clutter -1
0 0.25 B 0 15 A Firing rate Response 370 (spikes /s) Position FIG. 3 Fraction of
neurons Invariance -0.5 V1, mean: 0.02
IT, mean: 0.68 0 60 V1 simulated ORTICAL NEURONS COPE WITH CLUTTER? 0 IT Neurons in IT are not “Invariant”; they are “Tolerant” 7. Summary of the goodness of d
View Full Document
This note was uploaded on 09/08/2013 for the course BBB 217 taught by Professor Nicolerust during the Spring '12 term at UPenn.
- Spring '12