Olivier Marre

Enjoy the read !

A citation from 1744 (no typo !).

A reference where Von Neumann already asked whether connectomics was the best way to understand a neural system, in 1949 (!).

A discussion on the pros and cons of Maximally Exciting Images, and how to go beyond them.

We propose another way break down the problem of understanding natural scene processing. Instead of breaking down the natural stimulus into simpler pieces, breaking it down in different natural visual tasks.

We discuss the "curse of model complexity" and the issues that come with it.

We encapsulate both efficient coding and labeled line theory as specific cases in a more general framework (task-driven reductionism).

Other reasons why you may want to read it:

In this paper we focus on emerging approaches to characterize retinal processing of natural scenes, many of them being relevant beyond the retina. We have discussed many ideas there, and hope this will be interesting for many.

Retinal processing of natural scenes: challenges ahead While substantial knowledge exists about the way the retina processes simple stimuli, our understanding of how the retina processes natural stimuli re…

How does our visual system process natural scenes ? How can we approach this question ?
Happy to share this recent review written with Samuele Virgili where we ask these questions at the level of the retina.

www.sciencedirect.com/science/arti...

Exposition Bridget Riley et Seurat au musée d’Orsay jusqu’au 26 janvier, l’artiste presque centenaire, gloire de l’art cinétique, eu la révélation des interactions des couleurs en copiant Seurat en 1959. L’occasion rare de faire l’expérience sensible des liens reliant ces 2 géants de la perception

This head is spinning continuously, but we see it rotating back and forth...

...presumably because of our strong prior expectation that faces are convex.

This is a very nice example of the Hollow-Face illusion promoted by Richard Gregory:

www.richardgregory.org/experiments/

Excellent work @victorcalbiagueg.bsky.social !!!

Zebrafish use spectral information to suppress the visual background Vertebrate eyes first evolved in water, where spectral content rapidly fades with distance. Zebrafish exploit this loss by antagonizing cone signals to suppress the background, pointing to distance es...

Excited to share our paper now published in Cell!
'Zebrafish use spectral information to suppress the visual background'

Huge thanks to @neurofishh.bsky.social & @teuler.bsky.social

@cellpress.bsky.social @cp-cell.bsky.social

👇🏻
www.cell.com/cell/fulltex...

This is really cool work! Congrats, Olivier, and the whole team!

There is more to do on this, but this is an interesting framework to explain results from studies aiming at slowing down myopia. Stay tuned for more !

Another consequence is that, if you change the spherical aberrations, you might misguide your retina and make the eye more myopic. Actually, near vision does exactly that, and it is an important cause for myopia.

Actually, this is what myopia control glasses are doing ! These glasses reshape the visual input to try to have the retina slow down the eye growth. They are based on various designs, but the common point is that they decrease the spatial contrast received by the retina.

First, if this hypothesis is true, then decreasing spatial contrast should “trick” the retina into “believing” the focal plane is in front, and to slow down the eye growth.

We thus propose a possible strategy for the retina to detect the sign of defocus and adjust the eye growth: compute spatial contrast, and it works because of spherical aberrations. This gives two predictions for myopia.

Why is spatial contrast informative ? Because of spherical aberrations in the eye optics. This made the positive and negative defocus different.

When asking what these ganglion cells were doing, we found they computed spatial contrast, i.e. the variance of light intensity inside their receptive fields.

We found some ganglion cells that were always decreasing their firing, no matter what the image, when switching from negative (behind the retina) to positive (in front) defocus.

For this we decompose the problem in the two components: the eye optics, that we simulated, and the retina, that we recorded. We displayed to the retina natural images transformed by the eye optics.

Why is spatial contrast informative ? Because of spherical aberrations in the eye optics. This made the positive and negative defocus different.

When asking what these ganglion cells were doing, we found they computed spatial contrast, i.e. the variance of light intensity inside their receptive fields.

We thus asked: what kind of computations would allow the retina to extract the sign of defocus ?

Based on this, several attempts have been made to ‘’trick’’ the retina into “believing” that the focal plane is in front, so that it slows down the eye growth. Results are interesting, but would probably benefit from knowing how the retina does that.

…our retina can ! If the image is focused in front of the retina, eye growth is slowed down to try to bring the focal plane into focus. Picture from Carr and Stell.

Can we do something about it ? Myopia is an excess of eye growth, and often progress during childhood and adolescence. Can we slow down this growth ? Well, it turns out that…