Lecture 5: An approach to vision The implementation of vision Flashcards
Experimental approach
Varied techniques for understanding internal mechanisms
mental images are:
analogical
Mental images lack inherent structure, making no automatic distinction between ___, __, and ___ —> additional analysis is essential.
objects, figure and ground
Mental images must be/are
coded
We compute
representations
Representations are
stable,
require/allow for updating
Representations also allow for
Reference, thinking and action
This theory of vision suggests that when a mental image is created, the process is inherently ____.
analogical
Mental Images and the Internal Eye:
The concept implies that viewing a mental image necessitates an “internal eye” to observe it, which leads to a recursive question—if the internal eye observes the image, does another internal eye observe that, and so on? This recursion hints at an infinite regression, challenging how we actually “see” mental images in the mind.
Indistinction in Images
Mental images don’t inherently differentiate between objects, figure, and ground (the background or context within an image). This absence of distinction means that for a mental image to be understood, further cognitive analysis is required to segment and interpret the different elements within it.
Need for Coding:
For mental images to be useful, they must be coded—essentially processed, interpreted, or transformed into information that the brain can meaningfully use. Coding allows the brain to make distinctions and draw insights, converting raw imagery into structured knowledge.
coding allows the brain to
make distinctions and draw insights, converting raw imagery into structured knowledge.
mental Images as Coded Representations:
Mental images are not just visual impressions; they are coded representations. This coding allows the brain to store and process these images in a structured form, enabling more complex mental operations than simple recall.
Computation of Representations:
The brain computes these representations, actively creating a stable but flexible model of the image. This computational process transforms sensory input into meaningful constructs that can be interpreted, manipulated, and referenced.
Stability and Updatability:
Representations in the mind are stable, meaning they can be retained over time. However, they also allow for updates, enabling new information to adjust or refine existing representations. This adaptability is essential for learning and responding to changing contexts.
Reference, Thinking, and Action:
Coded representations allow the mind not only to refer to these images (calling them to mind) but also to think about them and use them to inform decisions and actions. This functionality supports higher cognitive processes, where visual information can guide reasoning and behaviour.
Imperfect Visual Input
The images we perceive from the world are inherently imperfect. This imperfection could stem from physical limitations (like blind spots, variations in light, and occlusions) and the incomplete nature of what reaches our eyes.
what are physical limitations that can give rise to imperfect visual input
blind spots, variations in light, and occlusions
“Holes” in Images:
Due to these imperfections, there are “holes” or gaps in the visual information we receive. The brain doesn’t receive a complete, seamless picture; instead, it has to work with fragmented data and partial visual cues.
Compensation by the Visual System:
The brain’s visual system compensates for these gaps. Through sophisticated processes, it fills in missing details, makes sense of ambiguous shapes, and corrects distortions to create a coherent mental image. This compensatory function is crucial for interpreting the world in a way that feels whole and stable.
Transformations via Computations:
The process of compensating for these visual imperfections requires numerous transformations. These transformations—referred to as “computations”—involve analyzing, organizing, and adjusting raw sensory input to build a more reliable internal representation. These computations are essential for the brain to make sense of incomplete or flawed data.
Computational Theory:
This level focuses on understanding the purpose or objective of a given cognitive process. It defines what the computation aims to accomplish and why it’s necessary. For instance, in processes like attention, sentence processing, or object recognition, the goal might be to focus selectively on information, understand language, or identify objects in the visual field. This level clarifies the problem being solved or the function being carried out by the system.
Representation and Algorithm:
At this level, attention shifts to how the cognitive system achieves its goals. This involves specifying the types of REPRESENTATIONS (or coded forms of information) the system uses and the RULES or algorithms that manipulate these representations to achieve the intended outcomes. For example, different types of representations might be visual patterns, linguistic structures, or memory traces, each governed by rules specific to its domain.
Implementation:
The third level concerns the physical realization of the system. Here, the focus is on the actual physical or biological substrate that carries out the computations and implements the algorithms. In cognitive science, this usually refers to the BRAIN or other physical devices like COMPUTERS that host and perform these processes.
V1 has:
A full map of the retina
Rods:
Dim light vision, movement detection
Cones:
Fine details, colour vision
Sequence from eye to V1 and beyond
1.Reception
2.Transduction
3.Coding
4.Representations and Processes
Reception:
Absorption of physical energy by the receptors
Transduction:
Physical energy is converted into an electrochemical pattern in the neurons
Coding:
One-to-one correspondence between aspects of the physical stimulus and aspects of the resultant nervous system activity
Representations & Processes
- What we do with what we encode from the world
- How vision affords knowledge and action
