Nostr notes by toerror

I found with height field data, it tends to smush together the ...

2026-08-08T13:42:18Z

In reply to nevent1q…gkx0
_________________________

I found with height field data, it tends to smush together the quantisation stair cases as if they were plasticine, and you pushed across them smoothing them out, but you can find it wrecks other details that you'd rather keep.

So "Bilateral" filtering is essentially an edge ...

2026-08-08T13:40:46Z

In reply to nevent1q…dc5d
_________________________

So "Bilateral" filtering is essentially an edge preserving blur. It performs a bunch of samples and integrates them together like a weighted Gaussian blur that preserves details.

Here's an example performing an image blur -
https://www.shadertoy.com/view/4dfGDH

Ah, fair enough - I ended up implementing bilateral to get rid of ...

2026-08-08T13:08:40Z

In reply to nevent1q…3j7l
_________________________

Ah, fair enough - I ended up implementing bilateral to get rid of exactly that quantisation, which it does with some success ( not complete removal ), but it does also attack desirable details to a slightly lesser degree. In the end I managed to get higher res data which removed the need, so if that's an option then that's the best approach. Worth remembering though if you find yourself stuck with display data.

It might be worth experimenting with bilateral smoothing of the ...

2026-08-08T11:47:46Z

In reply to nevent1q…zhqn
_________________________

It might be worth experimenting with bilateral smoothing of the height field data to try to smooth out some of those more flat regions so you don't get as much of that layering kind of effect. ( Apologies if you've tried this.. )

Yeah - I think there are different categories of these things - ...

2026-07-15T12:12:37Z

In reply to nevent1q…jfjf
_________________________

Yeah - I think there are different categories of these things - flat earth strikes me as a bit of a cult where membership has a requirement that you be willing to prove your devotion by rejecting reality, or the reality of those outside the cult. This also has the effect of alienating members from the outside world.

Off topic, but my theory is it's because they're ...

2026-07-15T12:01:16Z

In reply to nevent1q…ms7t
_________________________

Off topic, but my theory is it's because they're unprovable - or at least, it's the unprovable ones that stay around as conspiracy theories as opposed to established facts and therefore have longer to gain social inertia than other things which lose interest once they've been clarified.
Conspiracy theorists them selves appear to be attracted to the things that make them paranoid, reinforcing their anxiety surrounding the topic in question.

Inference costs aren't negligible at present though - or do ...

2026-07-02T09:06:11Z

In reply to nevent1q…5rmw
_________________________

Inference costs aren't negligible at present though - or do you mean for limited context single queries as opposed to e.g iterative vibe coding?

It looks like a project designed to fail after everyone concerned ...

2025-09-23T12:05:26Z

In reply to nevent1q…zcl5
_________________________

It looks like a project designed to fail after everyone concerned has got rich and retired.

They look like a horror concept.. I wonder if the original idea ...

2024-08-16T12:42:16Z

In reply to nevent1q…zm3m
_________________________

They look like a horror concept.. I wonder if the original idea was something to do with the creepy security risks of smart appliances, and someone got hold of the concept and recycled it out of context. It just seems like such a weirdly inappropriate idea to come up with out of the blue.