object recognition Flashcards
integration of visual dimensions into a coherent percept of an object
- lower levels of visual processing involves separate modules, that extract information about different dimensions
- the middle levels of visual processing begin recombining this information.
- higher levels of visual processing need to make sense of this information in a way that is independent of the viewpoint.
Marrs approach to understanding recognition
- Marr proposes that the first goal of making sense of the visual scene is to identify basic features, edge detectors. Our ability to distinguish the subject from the background is dependent on identifying the edge.
- Marrs edge detector’s: just like receptive fields of ganglion cells in the retina and LGN
- Output of LGN cell, at this level it is beginning to give a response. Burst of action potentials, there must be some sort of edge detection.
Past the LGN, PVC has cells which combine the edges,
- Output of LGN cell, at this level it is beginning to give a response. Burst of action potentials, there must be some sort of edge detection.
Curvature and shape detection in V4
- neurons in V4 code for surface shapes
- a viewer-centered representation of object. a template which must then be matched to other instances of templates
- not a complete representation of the object: it only exists in relation to the observer
object recognition in the Inferior temporal cortex
- Inferior temporal (IT) cortex is along the ventral path of the two visual streams model
- because of its role in object recognition, the ventral stream is also known as the ‘what ‘ pathway
- assumes there is hierarchal processing, although there is a lot of crosswork.
- large receptive fields: robust response even when object moves within the field, or changes in size, so not just a pattern of light the neurones responding to but also our understanding. n
seperate functional modules in the ventral stream
V4- analysis of form; processing of color constancy
LO- object recognition
FFA (fisiform face area)- face recognition, object recognition and experts
PPA (Parahippocampul place area): recognition of particular places
EBA- perception of body parts
visual object agnosia
- agnosia (failure to know)
- damage to the inferior part of the temporal cortex
- loss of the ability to recognize familiar objects
- IT is active when humans look at objects.
- difficulties depend on extent of lesion
variation in clinical cases
- Mrs R had a stroke that caused damage to the ventral stream of her extrastriata cortex
- can’t recognize a magazine, but can read
- can’t recognize facts- propagnsoia
patient JS had a stroke that cause extensive damage to the ventral stream - unable to recognize objects or faces
- could no longer read
- could not recognize shapes or orientation of objects
response of individual face neurons in non-human primates
masking of critical features, such as the mouth or eyes led to a substantial but not complete reduction of response- cells respond to intact face not indv features
D. scrambling the parts of face eliminates the response.
clinical evidence for distinct systems of visual agnosia
- Farrah (1990) reviewed a large number of agnosia studies
- argued for two types
- structural mechanisms (based on identifying features)
- holistic mechanisms (based on identifying configurations)
response of these neurons to non-face stimuli
– cells are not responsive to individual features presented in isolation
- contrast is important to the FFA
adjacent subregions
- Ball et al., found that in both the human and the monkey brain, regions responded to faces and body parts were adjacent to each other, as were those that responded to objects and scenes of places
face blindness (prosopagnosia)
- 85 yr old unable to recognize famous faces, his wifes or is owns, but recognition of other objects was normal
- right hem dam
memory of scenes
- people viewed 10,000 scenes for a few seconds each: people could determine which of two images had been seen with 83%.
- problems: images were very distinct, enables a sparese representation
long term memory for object details
- observers often fail to notice large changes to objects within visual scenes
- common belief that we are subject to change blindness
but: are visual long-term memory representations really that lacking in detail?
