The Cones and rods are not evenly or precisely sensitized on the retina by incoming light. Because of this, the image is formed with distortions. But the brain corrects these deformations through a calculation that tries to form a uniform image, which is the end product we see. I would like to know if there is any algorithm that has already proposed to model how this probabilistic correction that the brain makes.

Does anyone know anything about this? It would be amazing


1 Answer 1


The brain works as a hierarchical distributed network of networks (hence heuristics).
We attach meaning and prototype of objects as abstract concept.

DNN was created to reconstruct images using fMRI.

But this is not the whole story, you have to know three additional things to make a connection.
Various networks or blending functions are individual on person basis. Everything starts from the first day of life.
We try our senses, here touch and vision. We see 3D, but to do that, we would require 3 eyes. So depth of image is also created in kind of brains distributed DNN using additional info, like shade or object recognition (cognitive function tells us that e.g. car has depth, so this part is purely cognitive).

The distribution of rods and cones is not uniform and not the same among people, but yet, we see the same things?
This is not entirely true, people who never seen buildings (experiment on native tribes) when go up, to the roof, they do not feel anxiety looking down.
People who never seem triangle are immune to optical illusion, two triangle illusion They see three bitten apples and pictogram of three birds.

Getting back, people who share knowledge, say starting from primary school, will see it.

The missing part: how people can see the same things with the different distribution of rodes and cones?
First, they see small differences, but are able to infer it. Second, crucial point, when we learn to use sight, we also use touch as training set. Being able to touch something and remember shape, calibrates perception. Based on other senses and continuous training (brain consolidates every new piece of information all the time), the distribution of sensors doesn't matter, as we check out the world and surroundings for something like first five years, and then start to infer and learn abstract ideas, with calibrated sight.

This is solely based on senses and experience.

This is the same as with neural network classifier that gets different modified versions of the same image, as far as it is used consistently the same convolution function, the results won't change much. People change this convolution function in the first years of life after touching enough objects and byproduct is depth sensing.

  • $\begingroup$ As you mentioned there, the brain does not only interpret the environment through visual stimuli. In addition to the information provided via vision, the brain uses its internal "database", which begins to be built from birth (and which never stops) by sight and other senses, such as touch. It is likely then that the individual will develop, with more accurate perception of the environment, space, volume, texture and so on. stay. In fact, $\endgroup$
    – Octupos
    Commented Oct 4, 2021 at 11:58
  • $\begingroup$ Ruff et al. (1992) $\endgroup$
    – Octupos
    Commented Oct 4, 2021 at 11:58
  • $\begingroup$ Another point that I found interesting is the fact that even stereopsis (the perception of depth produced by the reception in the brain of visual stimuli from both eyes in combination; binocular vision) is a fundamental characteristic for our perception of space, even if if you lose vision on one side, the brain still manages to maintain good depth perception. And this, in part, is possible precisely because of the internal "base date" that the brain constitutes and continues to consolidate throughout life $\endgroup$
    – Octupos
    Commented Oct 4, 2021 at 11:59
  • $\begingroup$ Yes, you are right, Ruff et all showed this database for depth. Less sophisticated results are from times, when Eye Trackers were needles sticked into eyes, testing pilots perception, so prior to 1940. First non-intrusive Eye-Tracker was built in mid sixties. The first Eye-Tracker overall was in 1800, showing that we do not even read whole words. It was not apparent at the moment, but result is obvious now, people do not use sight 100% for vision, a huge part is cognitive database, $\endgroup$
    – Evil
    Commented Oct 4, 2021 at 13:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.