object recognition Flashcards
examples of object recognition in nature
echolocation
- e.g bats finding water = smooth surface targetted, rough surface avoided
electroreception
- fish create electric field and detect with electroreceptors and can detect any objects nearby
what is orientation invariance
being able to identify an object when it is at a different angle
what is scale invariance
being able to identify an object when it is a different size
heirarchical model of object recognition
categorisation
view-invariant object detection
detection of objects from one perspective
detection of object parts
detection of combination of edges and contours
detection of edges
lesions where decrease ability to recognise objects?
Lesions in inferior temporal cortex
role of lateral geniculate nucleus
sends information to primary visual cortex at back of brain
structure of lateral geniculate nucleus
- 6 layers
- recieve input from parvocellular (resolution) and magnocellular cells (motion)
- first 4 = PC, last 2 = MC
- layers recieve different inputs from contra and ipsilateral eyes
what brain structure is linked to the ventral visual pathway
inferior temporal lobe
key structural features of LGN
- 6 layers of neurones with different functions
- columnar structure allows occular dominance
- orientation and direction
- blobs
what are blobs
- parts of brain with specific properties
- contain neurones processing information about colour
- receive info from parvocellular cells of LGN
what is the ventral pathway
figures out WHAT the object is
information that leads to the temporal lobe, which is involved with object and visual identification and recognition.
what is the dorsal pathway
finds out WHERE the object is
leads to the parietal lobe, which is involved with processing the object’s spatial location relative to the viewer and with speech repetition.
what is lateral geniculate nucleus
structure in thalamas
what layers of LGN are magnocellular and parvocellular cells?
The top four are parvocellular layers, two layers from each eye. Parvo (small) LGN cells receive inputs from (small) midget ganglion cells.
The bottom two are magnocellular layers, one layer from each eye. Magno (large) LGN cells receive inputs from (large) parasol ganglion cells
where do P and M ganglion cells send their axons?
Retinal P ganglion cells send axons to a parvocellular layer, M ganglion cells send axons to a magnocellular layer
which eyes send info to which layers of LGN?
the eye on the same side (the ipsilateral eye) sends information to layers 2, 3 and 5
the eye on the opposite side (the contralateral eye) sends information to layers 1, 4 and 6.
areas sending info to LGN
retina, visual cortex, superior colliculus, pretectum, thalamic reticular nuclei, and local LGN interneurons.
what is centre-surround organisation
- Characteristic found in retinal ganglion and bipolar cells where the centre sends one type of signal, and the ring around the centre sends another.
- ON stimulates neuronal responses and OFF supresses neural responses.
role of the primary visual cortex
- LGN sends most of its axons there so it is the first processing area in cortex
- once V1 processes LGN info, it sends output to other visual cortices e.g V2,V3
- located in calcarine sulcus in medial occipital lobe
organisation of primary visual cortex
- columnar orientation
- columns respond to different stimuli
- chunks of cortex about 1 mm square by 3 mm thick that contains neurons, all with same receptive field location, but with all different orientation selectivities, direction selectivities, both (left- and right-) eye dominances represented.
what determines orientation of receptive field in simple cells
- Centres of the receptive fields of all neurons projecting to the same simple cell are oriented along a line
- Orientation of this line determines the orientation of the receptive field of the simple cell
how are receptive fields of simple cells elongated
each simple cell sums inputs from LGN neurones to build elongated receptive field responsive to bars and edges
features of simple cell receptive fields
- respond best to elongated bars or edges
- orientation selective
- separate ON and OFF subregions
- length summation
features of complex cell receptive fields
- orientation selective
- no separate ON/OFF regions
- binocular
- length summation
hypercomplex/end stopped cells
larger receptive field that fires when objects are anywhere in region
if bar goes away from receptive field, stimulates inhibitory cells
how are extremely complex shapes processed?
- grandmother/jennifer aniston neurones
- Neurone/group of neurones responds to images of a particular object (subject) in any orientation and scale
issues with hierarchical model of object recognition
- Still poor in scale- and orientation invariance
- doesnt take into account feedback from higher cortical areas
- needs experimental validation which is difficult because you need to record pre and postsynaptic responses