PSY1002 Lecture 2 2024 (Perception 1: Vision) PDF

PSY1002 - Lecture 2 Perception 1: Vision Daniel Poole Content notes: Description of animal experiments using invasive techniques for recording  Learning Objectives & Key Concepts Learning objectives Key concepts Explain the basic concepts of Sensat...

PSY1002 - Lecture 2 Perception 1: Vision Daniel Poole Content notes: Description of animal experiments using invasive techniques for recording  Learning Objectives & Key Concepts Learning objectives Key concepts Explain the basic concepts of Sensation, Perception, sensation and perception Receptors, Nerves Recognize and apply Gestalt Laws Gestalt Psychology Understand feature representation in Receptive field, visual the visual pathway hierarchy, edge detection Describe the two streams of visual Ventral and dorsal pathways processing Explain how bottom-up and top-down Cues, features, illusions processing combine to enable visual perception Reading McBride book, chapter 3 Last week recap Why Perception? At te nt io Perce n ption Languagieng Memor son y Rea Sensation & Perception  Sensation & Perception Sensation: the passive process of bringing information from the outside world into the body and to the brain. Perception: the active process of selecting, organising, and interpreting the information brought to the brain by the senses.  Sensation Sense Receptor Nerve conduit Brain organ➡️ cells➡️ ➡️ area  Sensation Sense Receptor Nerve conduit Brain organ➡️ cells➡️ ➡️ area Eye ➡️ Rods and cones ➡️ Optic Nerve ➡️ Visual Cortex Perception Approaches to study perception: Gestalt psychology Computational approach: feature representation Gestalt laws Gestalt psychology Gestalt: German for ‘form’ or ‘shape’ Paradigm which emerged in late 19th century Concerned with identifying laws which govern visual perception Gestalt laws Gestalt laws/principles: explain how parts are arranged into forms and objects, and perceived as a whole Gestalt laws Gestalt laws/principles: explain how parts are arranged into forms and objects, and perceived as a whole Gestalt laws Gestalt laws/principles: explain how parts are arranged into forms and objects, and perceived as a whole Gestalt laws Gestalt laws/principles: explain how parts are arranged into forms and objects, and perceived as a whole Dog Background Descriptive not mechanistic A major focus of Gestalt laws is describing the conditions that lead to grouping Gestalt psychology is less about understanding the mechanism underlying grouping, Which brain process leads to these rules.  Similarity Elements that look similar will be perceived as being part of the same form.  Similarity Elements that look similar will be perceived as being part of the same form.  Proximity Elements that are close together will be perceived as belonging together  Proximity Elements that are close together will be perceived as belonging together  Good continuation We perceive lines as following a smooth course. E.g. A - C and B - D appear as continuous lines  Closure A boundary isn't necessary for us to perceive a shape  Closure A boundary isn't necessary for us to perceive a shape When small elements are arranged in groups we tend to perceive them as larger figures  Closure A boundary isn't necessary for us to perceive a shape When small elements are arranged in groups we tend to perceive them as larger figures This can lead to us seeing illusory lines that do not exist Closure  Prägnanz (Simplicity) We organise a scene according to its simplest (shortest) explanation. Is this scientific? How do we decide what the simplest explanation is – i.e. precisely what makes one scene simpler than another? Other Gestalt laws Common fate: Elements that move together tend to be grouped together. Other Gestalt laws Common fate: Elements that move together tend to be grouped together. Symmetry: Elements that are symmetrical tend to be grouped together. Other Gestalt laws Common fate: Elements that move together tend to be grouped together. Symmetry: Elements that are symmetrical tend to be grouped together. Parallelism: Elements that are parallel tend to be grouped together.  Are Gestalt laws still relevant? Measure responses when systematically varying by a Gestalt Law E.g. symmetry  Are Gestalt laws still relevant? Influential with designers For instance User Experience (UX) design Feature representation in the visual pathway Feature representation in the visual pathway Eye ➡️ Rods and cones ➡️ Optic Nerve ➡️ Visual Cortex The visual pathway 1.Retina 2.Optic nerve The visual pathway 1. Retina 2. Optic nerve 3.Thalamus The visual pathway 1. Retina 2. Optic nerve 3. Thalamus 4.Primary visual cortex (V1) The visual pathway 1. Retina 2. Optic nerve 3. Thalamus 4. Primary visual cortex (V1) 5.Higher visual cortices (V2, IT) The visual pathway 1. Retina 2. Optic nerve 3. Thalamus 4. Primary visual (striate) cortex (V1) 5. Higher visual cortices (V2, IT) How is information represented in individual neurons within each of these different areas of the visual pathway?  Investigating feature representation Idea: Record from a single neuron, present different visual stimuli, check which ones elicit a response  Investigating feature representation Idea: Record from a single neuron, present different visual stimuli, check which ones elicit a response  Investigating feature representation Idea: Record from a single neuron, present different visual stimuli, check which ones elicit a response no response  Investigating feature representation Idea: Record from a single neuron, present different visual stimuli, check which ones elicit a response no response no response  Investigating feature representation Idea: Record from a single neuron, present different visual stimuli, check which ones elicit a response no response no response high response Feature representation in the optic nerve and thalamus Feature representation in the optic nerve and thalamus Receptive fields (features that neurons are most responsive to): Feature representation in the optic nerve and thalamus Receptive fields (features that neurons are most responsive to): Feature representation in the optic nerve and thalamus Receptive fields (features that neurons are most responsive to): Centre-surround organisation Light centre and dark surround or vice versa -> Neurons responsive to dot-like circular visual stimuli Feature representation in primary visual cortex Feature representation in primary visual cortex Classic studies by Hubel and Wiesel in 1950s investigating primary visual cortex (V1) Neurons in V1 respond to circular stimuli, but are not very active Feature representation in primary visual cortex Classic studies by Hubel and Wiesel in 1950s investigating primary visual cortex (V1) Neurons in V1 respond to circular stimuli, but are not very active Instead, they are most excited by line stimuli of a specific orientation Edge detection Feature representation in primary visual cortex Feature representation in primary visual cortex Feature representation in primary visual cortex Feature representation in primary visual cortex Feature representation in primary visual cortex  Feature representation in primary visual cortex Thalamus Receptive fields (x5 neurons) Visual cortex Receptive fields (x1 neuron) Feature representation in primary visual cortex Receptive fields in V1 are built up by combining receptive fields of neurons in the thalamus Feature representation in primary visual cortex Receptive fields in V1 are built up by combining receptive fields of neurons in the thalamus Hierarchical processing Feature representation in primary visual cortex Receptive fields in V1 are built up by combining receptive fields of neurons in the thalamus Hierarchical processing Receptive fields get larger as one ascends the hierarchy Feature representation in primary visual cortex Receptive fields in V1 are built up by combining receptive fields of neurons in the thalamus Hierarchical processing Receptive fields get larger as one ascends the hierarchy Receptive fields get more complex Activity Discuss with people sat near to you: What does it mean that feature perception is organised hierarchically?  Visual feature processing processing Bottom-up  Visual feature processing processing Bottom-up Optic nerve / Thalamus  Visual feature processing processing Bottom-up Primary visual cortex (V1) Optic nerve / Thalamus  Visual feature processing Secondary visual cortex (V2) processing Bottom-up Primary visual cortex (V1) Optic nerve / Thalamus  Visual feature processing Inferior temporal cortex (IT) Secondary visual cortex (V2) processing Bottom-up Primary visual cortex (V1) Optic nerve / Thalamus Break What has been unclear so far? Sign in! Two streams of visual processing  Two streams of visual processing Ungerleider & Mishkin, 1982  Two streams of visual processing “What” pathway Shape, objects Dorsal Stream – Where? ventral, to inferior temporal lobe Ventral Stream – What?  Two streams of visual processing “What” pathway Shape, objects Dorsal Stream – Where? ventral, to inferior temporal lobe “Where” pathway Motion dorsal, to superior parietal lobe Ventral Stream – What? Two streams of visual processing Simplified, but helpful view True picture likely very complicated Felleman & van Essen, Cereb Cortex, 1 Top-down vs bottom-up processing  How to think about visual perception? Computational Gestalt Psychology We reconstruct a We combine elements in visual scene by ways to gain a holistic combining simpler understanding of a elements scene  How to think about visual perception? Computational Gestalt Psychology We reconstruct a visual We combine elements scene by combining in ways to gain a simpler elements holistic understanding of a scene  How to think about visual perception? Computational Gestalt Psychology We reconstruct a visual We combine elements in scene by combining ways to gain a holistic simpler elements understanding of a Faithful scene reconstruction Simplified interpretation → built from visual → built using assumptions inputs alone and knowledge about world  How to think about visual perception? Computational Gestalt Psychology We reconstruct a visual We combine elements in scene by combining ways to gain a holistic simpler elements understanding of a Faithful reconstruction scene Simplified → built from visual inputs alone interpretation → built using assumptions and knowledge about world  How to think about visual perception? Computational Gestalt Psychology Computational approach Top-down vs bottom-up Gestalt approach  Top-down vs bottom-up High level: Associative brain areas Computational approach Gestalt approach Low level: visual inputs  Top-down vs bottom-up High level: Associative brain areas Computational approach Gestalt approach Bottom Up: processing the stimuli influences what is perceived → data driven Low level: visual inputs  Top-down vs bottom-up High level: Associative brain areas Computational approach Top Down: background knowledge and Gestalt approach expectations influence what is perceived → expectation driven Bottom Up: processing the stimuli influences what is perceived → data driven Low level: visual inputs When is top down information important when processing a visual scene?  Top-down information: Context matters Our environment gives us clues when the stimulus is ambiguous  Top-down information: Context matters Our environment gives us clues when the stimulus is ambiguous The context we perceive a stimulus in can change our perception of it.  Top-down information: Context matters Our environment gives us clues when the stimulus is ambiguous The context we perceive a stimulus in can change our perception of it. Visual illusions Many illusions are consequence of bottom-up vs top- down processing Assumptions can cause an “incorrect percept” (top- down) Illusions are useful as they allow us to access the assumptions  Size Contrast E.g. Ebbinghaus illusion The orange dots are the same size!  Size Contrast E.g. Ebbinghaus illusion The orange dots are the same size! Size contrast between the central and surrounding circles  Size Contrast E.g. 2. Deboef illusion  Size Contrast E.g. 2. Deboef illusion Light from above Expectations of rules about the world shape perception E.g. circles perceived as convex or concave bumped depending on shading Light from above Expectations of rules about the world shape perception E.g. 2  How do we resolve ambiguities? We need to decide which visual scene caused the image on the retina How we do this: Assumptions and Cues are combined to make a best guess at what it is we’re seeing Cues: features of the image that give clues as to the nature of the stimulus (bottom-up) Assumptions: expectations about what we will see or what different cues mean (top-down) Summary Bottom-up Computational approach (feature representation) Two streams of visual processing Top-down Gestalt Laws Context and assumptions Q&A