<othx>
peeps[win] linked to "Scripted vases by hakalan" on thingiverse => 1 IRC mentions
<ndnihil>
pretty sure the psu is about to eat itself
<linext>
are you still using a computer PSU?
<ndnihil>
or the rambo is about to catch fire
<ndnihil>
yeah
<ndnihil>
and not even an upgraded one
<ndnihil>
the one that came with it
<linext>
yea 400watt
<ndnihil>
it wheezes every time I turn it on
<peeps[win]>
oops, i gues that's a python script, not openscad based. i printed some of them so long ago I forgot
<linext>
chuck it and put in a 50amp 12v power supply
<ndnihil>
begs me to put it out of its misery
<SD210>
im bout to break into my uni super computer lab just to render some stl's lmao
<linext>
my heated bed and extruder heats up in like 2 minutes with 50a 12v
<linext>
sd210: if you install openscad with "manifold" it'll be fast
<ndnihil>
it dying is what made me start building this voron
<SD210>
thx
<ndnihil>
decided that things had progressed enough in the ~10 years I've been abusing that thing that it might be time to upgrade
<linext>
yea, input shaping will be nice. a friend just got his prusa i3 mk4 w/ mmu at my suggestion
<linext>
he decided to get the kit and is said assembly is no joke
<ndnihil>
I have plans for the rostock once I'm done with this voron
<ndnihil>
going to take the hotend off and stick my little tritium capsule on it, put it in a dark room, and do long exposure photography from multiple angles while it draws things in mid air
<dTal>
ooh la la
<ndnihil>
then photogrammatry that into a model
<ndnihil>
and slice it up for this little holocube thing
SD210 has quit [Quit: Client closed]
peeps[win] has quit [Read error: Connection reset by peer]
<ndnihil>
s/photogrammatry/photogrammetry/
<dTal>
what's the point of doing all that? you can just generate the model directly from the gcode you'd use to drive the tritium capsule in the first place
<ndnihil>
can't type for shit tonight
<dTal>
you're gonna program trajectories and then pass them through a photogrammetry pipeline because... ?
<ndnihil>
because I can?
<dTal>
but why
<ndnihil>
why wouldn't I?
<ndnihil>
I've already fed this thing models
<dTal>
because it's a time consuming and fiddly way to accomplish nothing
peeps[win] has joined #openscad
<dTal>
I mean I'm all for crazy artistic pipelines, you've got some cool pieces there
<dTal>
but that specific pipeline just takes some data and gives it the runaround without altering it artistically in any way
fling has quit [Remote host closed the connection]
<dTal>
Are you aware of neural radiance fields? It seems to be like for your application you want to keep the data in the image domain, and not try and triangulate anything
<ndnihil>
well, doing the model is just the first part of it
<ndnihil>
next step is controlling it with my brain
<ndnihil>
to make the drawings
<dTal>
what you might consider doing is putting the camera on the hotend instead of the tritium, and fixing the tritium somewhere. Then you can repeat the pattern as many times as you want but with translational offsets. This simulates many long exposures from different positions
<ndnihil>
already have my eeg assembled and working
<ndnihil>
nah, I'm fine doing it the way I had planned
<dTal>
how is that?
<ndnihil>
exactly as I described
<dTal>
you said you'd do a bunch of long exposures from different angles
<ndnihil>
multiple cameras
<dTal>
how many?
<ndnihil>
well, I have 6 identical camera modules
<ndnihil>
so maybe all of them
<dTal>
You need more like 50-100 for photogrammetry / NeRF
<ndnihil>
wide'ish angle, so it'll give decent coverage
<ndnihil>
dude, I've been doing dumb shit like this for decades
<ndnihil>
I really don't need much in the way of advice
<ndnihil>
sub-mm accurate 3d model of one of my fossils
<dTal>
I have a lightfield scanner which is a creality cr-10s with the hotend swapped for a camera, and I plan to build a lightfield camera with 100 or so camera modules triggered simultaneously
<dTal>
and hopefully, maybe, a holographic printer to make hard copies
<ndnihil>
eh, lightfield is novel, and kinda neat, but shit for practical
<ndnihil>
ask lutron or lytron or whatever they called themselves
<dTal>
lytro
<ndnihil>
yeah them
<dTal>
don't be shitting on lightfields
<dTal>
it's gonna change things, radically
<ndnihil>
it's an insanely interesting technology
<dTal>
*especially* VR
<dTal>
lytro were hacks
<ndnihil>
and "they" have been saying it'll radically change things for what, 10-15 years now?
<dTal>
no
<dTal>
things are different now
<ndnihil>
beyond just the sensors getting better?
<dTal>
The sensors aren't much better, what's better is the software, and yeah that changes everything
<ndnihil>
if applied properly, sure
<dTal>
we also have consumer VR, that's a big deal
<ndnihil>
but whoever is going to give you funding is going to be the same entity that chokes your new tech to the point nobody wants to use it
<dTal>
with a little application, you can now port a physical environment into VR, photorealistically
<dTal>
there's demos you can try right now
<dTal>
whatever you've seen in VR so far, it's a completely different ball game
<ndnihil>
demos that wont make you get a headache and/or get nauseous?
LordOfBikes has joined #openscad
<ndnihil>
one you cross that uncanny valley, your brain starts to get -real- irritated with inconsistencies
<dTal>
you can't render scenes like that in real time using traditional techniques, but this is basically just a retrieval / interpolation task
<ndnihil>
s/one/once/
<dTal>
there was a paper at siggraph only a few weeks ago, anisotropic gaussian splatting - basically, lightfield rendering in real time, off data structures than can be trained using gradient descent from a few dozen photos
<dTal>
this is *huge*
<ndnihil>
sounds cool for sure
<ndnihil>
but the older I get, the less excited I get about "sounds cool" before there's something practical I can play with
<dTal>
it's practical now
<dTal>
for instance you wanna do photogrammetry
<dTal>
I know you didn't ask for advice but to me this sounds like a wayyy better fit for radiance fields (learning some visual scene representation through gradient descent)
<dTal>
because you don't want a model, in the end, you want a bunch of rendered images from different angles - right?
<dTal>
you're not 3d printing, you're displaying on a screen
<dTal>
and there's all sorts of nifty optical effects you'd quite like to keep
<ndnihil>
I'm entertaining myself with a novelty
<ndnihil>
I dont really have any sort of grand plan for it
<ndnihil>
just things to keep me engaged
<ndnihil>
mental fidget spinner
<dTal>
if you want to take a few dozen photos of a scene, and generate arbitrary new viewpoints, this is state of the art
<dTal>
well I'd love to see what you come up with, anyway
marcus has quit [Remote host closed the connection]
<ndnihil>
I'm sure I'll end up spamming "hey look what I did!" links to anyone who'll listen
<dTal>
long exposure lightfields though, damn. Didn't think of that
<ndnihil>
sounds like fun
<ndnihil>
if you have the hardware
<ndnihil>
and the software
<dTal>
planning on using ov5640 modules
marcus has joined #openscad
<dTal>
been thinking about things like burst mode, video etc
<ndnihil>
write the software for the leenooks
<ndnihil>
and I'll play
<dTal>
trouble is they'll be driven by esp32 modules with 4mb of ram, which doesn't exactly let you store a lot of frames
<ndnihil>
yeah they're fun, but a bit limited for things
<ndnihil>
this little holocube is esp32 driven
<dTal>
in fact it doesn't even let you store a single raw frame from an ov5640
<InPhase>
linext: You also have to factor in how much labor you save by not having to ship out individual units yourself. It's not purely about the profit per unit, but about profit per time spent if you can scale up to larger volume.
<dTal>
well, the looking glass works very nicely on linux :) even has a neat blender plugin
<ndnihil>
fucking blender...
<dTal>
but this is cool
<ndnihil>
it used to be such a great modeling program
<InPhase>
linext: Given the number of such knives out there, you surely aren't saturating the potential market. You're just not really reaching it. :)
<ndnihil>
then it turned into the worlds most annoying animated movie studio
<dTal>
the marketing copy on that tilt5 thing is obnoxious
<dTal>
"What AR u here to do?"
<dTal>
one thing the looking glass gets very right is software
<ndnihil>
eh, jeri isn't doing all of it
<ndnihil>
she's legit though
<dTal>
a product like that lives or dies by the SDK
<ndnihil>
backed the first iteration "castar" project way back
<ndnihil>
it's unity fed, they just haven't released the linux bits last I checked
<ndnihil>
if it had been anyone but jeri I probably wouldn't have backed the project
<ndnihil>
but girl welds race cars and has her own SEM
<ndnihil>
and designs FPGAs
<dTal>
not really a fan of the unity thing
<dTal>
it's gotta be more like VR
<dTal>
I know that's a tall hill to climb
<dTal>
but imagine if all VR games had to be unity
<dTal>
oh okay there's a "native C SDK" whatever that it
<linext>
yea, i have to dedicate a day to getting listed on Amazon
hyvoid has quit [Remote host closed the connection]
rue_mohr has joined #openscad
<rue_mohr>
is anyone working on path_extrude ?
snaked has quit [Ping timeout: 244 seconds]
<J23k13>
i have a module doing it, and BOSL or DotCAD has it too
J23k13 has quit [Quit: Client closed]
J23k13 has joined #openscad
<J23k13>
ndnihil are you planning to print and use your "eccentric cycloidal drive" ? because it will probably have some wear as the profile is not calculated as a cycloide