Show HN: High-Res Neural Cellular Automata

(cells2pixels.github.io)

126 points | by esychology 5 hours ago

8 comments

  • whilenot-dev 2 hours 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

    • esychology 2 hours 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 hour ago
        I would love to see two seeds competing for space in the grid
    • WhiteNoiz3 1 hour 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.
      • oersted 15 minutes 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.

        • WhiteNoiz3 10 minutes ago
          There's hidden state in the model which presumably it uses to communicate position, ie there's the 3 colors but then a bunch of other channels that the model can use how it wants.
      • cl3misch 21 minutes 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!
        • WhiteNoiz3 9 minutes ago
          I have yes.. You need to erase a larger amount of the center, but it almost always results in a collapse wheras erasing around the center typically regrows.
        • Mithriil 12 minutes ago
          If you hold the eraser for a second at the center, I find that it destroys the image more often than not.
  • WhitneyLand 1 hour 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…

    • oersted 27 minutes 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.

  • hidelooktropic 1 hour ago
    For the unfamiliar, could someone explain what I'm looking at? The abstract was a little too concrete (heh) for me to follow.
    • soraki_soladead 38 minutes ago
      The original NCA is probably a helpful intro: https://distill.pub/2020/growing-ca/
    • esychology 1 hour 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).
      • flir 1 hour 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).
        • esychology 1 hour ago
          The input to the NN is just the 3x3 neighborhood around a cell. We can overlap two NNs on the same grid (through interpolation). Checkout https://meshnca.github.io to see the effect. When the brush is in graft mode, it basically allows you to paint some regions that will follow a different NN rule.
          • flir 1 hour ago
            > The input to the NN is just the 3x3 neighborhood around a cell.

            Well that sounds like black magic. Nice. Thanks for the reply.

  • jekude 3 hours 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?
    • esychology 1 hour 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-shape 3 hours 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 2 hours ago
      You can, just enable the 'transition' switch.
      • embedding-shape 25 minutes ago
        That seems to be something else? It takes the current image and "transforms" it into the new target.
  • WithinReason 3 hours ago
    You can make the centipede grow longer, which makes sense given how this works. Or grow a 2nd centipede for extra points.
    • esychology 3 hours ago
      haha yes, also the same with the worm
  • amelius 3 hours ago
    Why are the images always generated in the same orientation (upright)? Do the cells have awareness of what is "up"?
    • WhiteNoiz3 4 minutes 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.
    • esychology 3 hours ago
      yeah normally NCAs have a sense of up and left. There are some isotropic variants that make the perception fully rotation-invariant.
  • mirekrusin 2 hours ago
    So the goal is to evaporate it with minimum number of shots?