Object recognition

levels of visual processing

• lower - separate specialised modules extract different features

• middle - combines info from modules

• higher - makes sense of info in viewpoint-independent way = object recognition regardless of angle/position

contrast detection by retinal and LGN

based on light thru centre-surround receptive fields (ON/OFF cells)

• responds to change in contrast - lines, edges and bars

• no response to uniform surface

orientation coding

in V1 - simple cells detect specific orientations of edges and lines (vertical, horizontal) by combing input from LGN with aligned receptive fields

• horizontal receptive fields = responds to horizontal edges

• vertical fields = responds to vertical edges

curvature and shape detection

in V4 - combine input from V1 to detect curved lines and complex shapes. represented from view-dependent format

inferior temporal cortex

part of ventral stream - highly selective neurons support advanced object recognition

visual object agnosia

unilateral or bilateral damage to inferior temporal cortex - unable to recognise familar objects based on sight

• can still recognise thru other senses

2 types of agnosia

• structural mechanisms based on identifying features

• holistic mechanisms based on identifying configurations

face inversion effect

facial recognition is impairedwhen presented upside down = suggests there is special processing of facial features

thatcher illusion

changes to facial features appear normal when face is upside down but obvious when right side up

why is facial processing seen as holistic

neurons are not responsive to individual features presented in isolation

fusion face area

Area in the brain responsible for processing faces

• debate on whether its specialised for faces or just expert objects

prosopagnosia

damage to fusiform gyrus (in inferior temporal) = face blindness

Brady et al 2008

Participants asked to detect repeated images in a stream with 0 to 1000+ intervening items = requires high fidelity memory of each image

• able to detect repeats even after large delays with many intervening = large capacity visual long-term memory

parahippocampal place area

borders ventromedial temporal lobe = activated by sight of scenes and backgrounds

• doesn’t depend on recognition of objects within scene