Psychophysical and Physiological Evidence for Viewer-centered Object Representations in the Primate

1
Psychophysical and Physiological Evidence for
Viewer-centered Object Representations in the
Primate
N.K. Logothetis and J. Pauls
Cerebral Cortex (1995)
2
Background
Input representation
Image
Match ?
Recognition
Transformations
Stored memory representations
Theories of object representations
3D models (Marr, Biederman) View dependent 2D
templates (Basri Ullman, Poggio)
Face selective cells
Found in STS Mostly view dependent
3
Methods
Trained three juvenile rhesus macaques on an
object recognition task Performed psychophysical
tests after training Recorded from the upper
bank of the anterior medial temporal sulcus (AMTS)
Stimuli Computer generated wire like and
amoeboid objects
4
Training

Began with training monkeys to recognize a single
view of an object presented sequentially among
distractor objects Slowly increased rotations up
to or 90o before training with a new
object Feedback with juice reward
5
Testing
6
Recordings
Recorded from 773 neurons in AMTS
7
Findingspsychophysical
Recognition performance fell off sharply when
object rotated more than 30-40o beyond training
view Both for wire and amoeboid objects
8
Findingspsychophysical
Interpolation with wire objects
Monkeys could interpolate between two training
views up to 120o apart
Three to five views allowed monkey to generalize
to entire great circle
9
Findingspsychophysical
Pseudo-mirror symmetrical wire objects
Some of the wire objects have mirror symmetrical
0o and 180o views due to lack of self-occlusion
10
Findingspsychophysical
Viewpoint invariance for basic objects among
different class distractors
11
Findingsphysiological
View specific, object specific cells (71 of 773)
Cell responses to distractor views
Cell responses to target views
12
Findingsphysiological
View invariant, object specific cells (8 of 773)
13
Findingsphysiological
14
Findingsphysiological
Multiple cells tuned to different views of the
same object
15
Authors conclusions
Object recognition depends on training view A
small number of stored views can be used to
achieve invariance with wire like objects Neurons
in IT found that respond selectively to learned
objects, mostly to specific views
Problems
Highly unnatural stimuli View selective neurons
used for recognition or after recognition?
Interpolation with self occluded (solid) objects?