This is amazing. Now I'm imagining the ability to simulate less-than-ideal conditions for tree growth. Placing the tree next to a building or on a hill or culling branches at certain stages in its life to see what happens next, etc.
I think my overall thought is: we see plenty of nice looking trees in games and graphics. What about authentically degenerate trees?
Now I want a simulated caterpillar. In my area, we sometimes get caterpillar apocalypse where the whole forest will be full of them, sometimes they cover entire trees with their silk.
I think the most interesting result would be somewhere halfway between 0 and 100% nibbled.
Don't stop there: fungi and beetles are some other stressors that could be modeled alongside caterpillars using similar techniques. Like lepidopterans, they are very host-specific, though they have somewhat different transmission vectors, so you could see the defoliation (or other foliage impacts like leaf curling and yellowing) spread / not spread depending on factors such as proximity, moisture, wind and the parasite's lifecycle.
Pretty cool idea: "buds compete for light, the tree distributes a growth resource among them, and the successful ones produce new shoots. Those shoots change the light available for the next round. The crown shape emerges from this feedback loop."
I would never have thought to generate tree shapes by simulating real life
There has been a complete deluge of new accounts sometimes only hours or days old being extremely condescending, overly negative, sometimes antagonistic, and sometimes they seem like a kind of LLM prompted to cause arguments.
The generated trees look absolutely amazing. Adjusting the process for different lighting situations seems straightforward too - freestanding trees look different from those in woods.
All their growth models do involve light, however. So it probably does make sense to have a common light representation - ideally some sort of sparse cone storage though, not a voxel grid the size of a forest.
This is amazing. Now I'm imagining the ability to simulate less-than-ideal conditions for tree growth. Placing the tree next to a building or on a hill or culling branches at certain stages in its life to see what happens next, etc.
I think my overall thought is: we see plenty of nice looking trees in games and graphics. What about authentically degenerate trees?
Now I want a simulated caterpillar. In my area, we sometimes get caterpillar apocalypse where the whole forest will be full of them, sometimes they cover entire trees with their silk.
I think the most interesting result would be somewhere halfway between 0 and 100% nibbled.
Don't stop there: fungi and beetles are some other stressors that could be modeled alongside caterpillars using similar techniques. Like lepidopterans, they are very host-specific, though they have somewhat different transmission vectors, so you could see the defoliation (or other foliage impacts like leaf curling and yellowing) spread / not spread depending on factors such as proximity, moisture, wind and the parasite's lifecycle.
Pretty cool idea: "buds compete for light, the tree distributes a growth resource among them, and the successful ones produce new shoots. Those shoots change the light available for the next round. The crown shape emerges from this feedback loop."
I would never have thought to generate tree shapes by simulating real life
Procedural generation always looks magical until you realize it's just a bunch of simple rules fighting for resources.
That makes it even more magical to me.
reductionists being reductionists. I'm with you.
There has been a complete deluge of new accounts sometimes only hours or days old being extremely condescending, overly negative, sometimes antagonistic, and sometimes they seem like a kind of LLM prompted to cause arguments.
HN needs to re-introduce the rule where you can't post until the account is at least a few days old.
occam's razor at its finest
The algorithm is cool and the article is pleasing to read.
This makes me realise that the amount of wind affects the number of viable leaves.
The generated trees look absolutely amazing. Adjusting the process for different lighting situations seems straightforward too - freestanding trees look different from those in woods.
oh hey, very nice result! i've been experimenting with trees (using ai) as well - https://treesim.netlify.app/
This is super cool! Been looking into procedural tree generation for a project and that paper is a great starting point.
Oh hell yeah, a topic like this was on my bucket list for a while.
Very nice overview of procedural tree generation!
Could foresters use this?
Nice to see another article from Mr Arion Render, always a pleasure. And of course posted by ibobev, always worth watching his HN submissions.
Show me a smoothly animated demo of growth from a bud! It would look incredible
now i am wondering if that would correctly imitate crown shyness etc. When several trees grow close to each other
It can, but doing that for a whole forest would require a lot of voxels!
Yes, especially as not all trees are the same age, and the older trees were grown defined by even older trees that are now gone.
yeah, and them having different rules for growth given different species
All their growth models do involve light, however. So it probably does make sense to have a common light representation - ideally some sort of sparse cone storage though, not a voxel grid the size of a forest.
This is beautiful! Thank you for sharing!