Show HN: High-Res Neural Cellular Automata

https://cells2pixels.github.io/

by esychology • 1 month ago

208 54 1 month ago

cells2pixels.github.io

Neural CAs model self-organizing pattern formation.

Now they can generate patterns at HD resolution in real-time, enabled by turning each CA cell into a Neural Field.

Try 3 demos: grow a pattern from a seed (and damage it, it heals), synthesize PBR textures that can regenerate, or create 3D textures like clouds.

amelius• 1 month ago

Why are the images always generated in the same orientation (upright)? Do the cells have awareness of what is "up"?

esychology11• 1 month ago

yeah normally NCAs have a sense of up and left. There are some isotropic variants that make the perception fully rotation-invariant.

WhiteNoiz35• 1 month ago

IIRC training starts with the initial state and the end state, and the end state is always oriented the same way. It would be interesting to see what would happen if the end state was rotated randomly though I suspect it wouldn't work so well.

esychology11• 1 month ago

Checkout the Isotropic NCA blog from the Google Zurich team: https://google-research.github.io/self-organising-systems/is...

WithinReason• 1 month ago

You can make the centipede grow longer, which makes sense given how this works. Or grow a 2nd centipede for extra points.

esychology11• 1 month ago

haha yes, also the same with the worm

jekude• 1 month ago

The abstract implies that strictly local updates are a hinderance to high res, however i would have thought there would be an interesting way to get speed up gains from neighbor-only traffic on GPUs CAM-style. am i making that up?

esychology11• 1 month ago

I think performance is not the only issue for scaling to larger grids. CUDA Convolution implementation already utilizes coalescing to improve performance. The main bottleneck is that in larger grids, cells are further apart, and it takes more steps for them to be able to communicate.

embedding-shape2• 1 month ago

Really interesting demo, nicely done :) Would be fun if switching the "Target Image" when using the second brush mode in the Growing Demo didn't erase/reset the existing canvas, so we could "stamp" new things on top of other images. Small thing perhaps but I got sad when it disappeared when I wanted to merge a kitten on top of the chameleon but couldn't :(

bfmalky• 1 month ago

You can, just enable the 'transition' switch.

embedding-shape2• 1 month ago

That seems to be something else? It takes the current image and "transforms" it into the new target.

esychology11• 1 month ago

That's not possible in the current demo but this sounds like an interesting feature to work on and add!

whilenot-dev• 1 month ago

The automata just completely destroys the image if I draw too much over the stabilized image with the brush. 5 horizontal swipes are enough to destroy the kitty, is that to be expected?

EDIT: video here: https://imgur.com/a/ItZGd5X

esychology11• 1 month ago

The NeuralCA both generates and maintains the pattern. Because the NCA was not exposed to damage or erasure during training, its regeneration capability is a purely emergent phenomenon. However, this ability remains somewhat brittle, particularly when the central regions of the pattern are erased.

mackenney• 1 month ago

I would love to see two seeds competing for space in the grid

WhiteNoiz35• 1 month ago

With the old model (and I suspect this one too) it's trained to generate from a single 'seed' pixel in the center of the image. If you erase the center of the image, that's when it completely collapses.

cl3misch• 1 month ago

Have you actually tried that? If you specifically erase the center, the image does change a lot at first, but rebuilds itself eventually (albeit to a slightly different final state). It's uncanny how "biological" is feels!

oersted2• 1 month ago

It must be more general than that, otherwise the cells wouldn’t be able to repair their area if the damage came from the wrong direction (repair is not center-out).

The model generally learns to generate each pixel from its surroundings, even if the surroundings are partially missing.

mirekrusin• 1 month ago

So the goal is to evaporate it with minimum number of shots?

hidelooktropic2• 1 month ago

For the unfamiliar, could someone explain what I'm looking at? The abstract was a little too concrete (heh) for me to follow.

esychology11• 1 month ago

If you're familiar with CAs (e.g. Conway's Game of Life), you can think of a NeuralCA as a CA where the update rule is given by a neural network. Here we optimize the neural net weights so that it behaves a certain way (e.g. grow a lizard from a single seed).

flir2• 1 month ago

What are the inputs to the NN? The whole grid, or just nearby cells? What happens if two NNs overlap on the same grid? (Gonna go read the paper).

hidelooktropic2• 1 month ago

Wow. That's fascinating. Thanks for that explanation. So these images come to be consequentially from initial state and weights...

soraki_soladead• 1 month ago

The original NCA is probably a helpful intro: https://distill.pub/2020/growing-ca/

WhitneyLand• 1 month ago

At a glance it looks like it could be just iterative texture sampling.

The difference is when creating each pixel, there’s no coordinate to look up, instead it’s using only a set of rules like Conway’s game of life.

But the rules come from a neural network trained on the image, so… it’s kind of memorizing enough information to effectively do the same thing as texture sampling, but using only local information.

I’m sure I’m missing something about how it works or what makes it interesting…

oersted2• 1 month ago

To me, it is intriguing as a toy model for how cells are able to grow into complex tissue and organisms based only on local information, and how they are able to repair and recover harmed tissue.

Of course, this is as close to cells, as neurons from neural networks are to real neurons. And I have no idea what it could be applied to (inpainting/outpainting?), but it’s interesting as exploratory research.

menno-sh• 1 month ago

Oooh, this made it click for me. Thank you

esychology11• 1 month ago

Texture sampling retrieves pixels by coordinate, while NCA grows them from local rules with no global lookup. The weights are actually ~3× smaller than JPEG-compressed texture maps, so it's not just memorizing the image either. The more interesting part is that it's a dynamic process, you can damage the pattern while it forms and it repairs itself.

zcw100• 1 month ago

I've always loved the original work and it's nice to see they're still working on it. I've always wondered if there was a way to connect this back to infrastructure rather than images. Something you could run on a cluster and if portions of it failed it would regenerate the system.

Like a bio inspired Kubernettes.....Bionettes.

bjourne5• 1 month ago

I found your previous work here: https://distill.pub/2020/growing-ca/ For someone (like me) who wants to understand the basics its probably better. It's very well written.

moralestapia2• 1 month ago

@esychology this is phenomenal work, thank you so much for sharing it. I am working in a similar thing and might reach out about it soon.

Also, what's going on? Why would the community flag and kill this comment[1], from the creator itself. If you're jealous of what the guy built, take it elsewhere. HN will implode with that attitude.

1: https://news.ycombinator.com/item?id=48571171

esychology11• 1 month ago

Thank you for the kind comment! Please reach out, I'm happy to have a chat.

bjourne5• 1 month ago

It's the stupid spam filter going haywire: https://news.ycombinator.com/item?id=48321198

yorwba4• 1 month ago

If it says [flagged], a human pressed the "flag" button. Most likely because OP is commenting too much.

_menelaus• 1 month ago

Fascinating. A glimpse into the mathematics of life and living tissue. A toy example like this might herald self-healing machines, in the same way that the first perceptrons heralded Fable and nude deepfakes.

IvanLudvig• 1 month ago

incredible work! I love seeing progress in this field. I've recently built a WebGPU implementation of the original growing neural CAs: https://github.com/IvanLudvig/neural-ca-webgpu

EstanislaoStan• 1 month ago

Managed to get a stable loop with the worm!

Akimsa3• 1 month ago

That's really impressive i've been thinking that we have using technologies cause they were a breakthrough but we actual resources we can implement technollogy that was relegated time ago, like when implementing 3d there were many other until the one who was more consistent is how we visualize 3d today. Cause this looks pretty simillar to metaballs. An idea that is not longer that hard pursued but i just saw the similarity.

Show HN: High-Res Neural Cellular Automata

Related Stories