--- Log opened Sat May 24 00:00:52 2025
02:19 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
02:27 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Quit: Avoid fossil fuels and animal products. Have no/fewer children. Protest, elect sane politicians. Invest ecologically.]
02:27 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
02:52 -!- stotznet [~stotznet@user/stotznet] has joined #hplusroadmap
02:55 -!- stotznet [~stotznet@user/stotznet] has quit [Quit: Leaving]
03:30 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
03:30 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
03:46 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Ping timeout: 264 seconds]
04:07 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
04:11 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Remote host closed the connection]
04:12 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
04:58 -!- yorick [~yorick@user/yorick] has quit [Quit: WeeChat 4.4.3]
05:00 -!- yorick [~yorick@user/yorick] has joined #hplusroadmap
05:44 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
05:44 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
08:49 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 260 seconds]
08:54 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
09:31 < hprmbridge> nmz787> So going back to 3D model generation from images.... There's a playground that's getting demolished this summer and I would like to capture it as a model, because it's a really cool design. Lots of bridges and twisting staircases and nooks and crannies to hide. I'm guessing I need as many photos as possible, and a video won't work. Probably a tripod would be ideal. And the software stack/tool(s)
09:31 < hprmbridge> nmz787> to try ?
09:34 < hprmbridge> nmz787> Seems no one suggested anything after I last mentioned meshy.ai (anything for image to 3D conversion)
09:34 < L29Ah> why won't video work?
09:35 < L29Ah> i think video coupled with DGPS location info is the best option for near-future technology
09:48 < TMA> nmz787_: https://zju3dv.github.io/neuralrecon/ https://openaccess.thecvf.com/content/CVPR2024/papers/Wu_PanoRecon_Real-Time_Panoptic_3D_Reconstruction_from_Monocular_Video_CVPR_2024_paper.pdf
09:58 -!- gptpaste [~x@193.29.58.204] has quit [Ping timeout: 248 seconds]
10:12 < hprmbridge> nmz787> Video blur was my assumption
10:13 < hprmbridge> nmz787> Academics are great, but "ain't nobody got time for that"
10:13 < hprmbridge> nmz787> (where I'm the nobody in this case)
10:17 < hprmbridge> nmz787> At the least just capturing the photos before the structure is destroyed is the minimum
10:19 < fleeky> TMA: oo thanks
10:19 < fleeky> i desperately need to get a decent scan of my face , maybe neuralrecon could do that hrmmm
10:33 < TMA> fleeky: if you would manage to use it, will you please tell me about your journey? I have read it but I have not got around to actually try it
10:38 < TMA> nmz787_: it seems that it is better to go overboard with the number of photos. I remember reading somewhere that for each reconstructed point at least 8 different photos are necessary and that was with the assumption that the location and angle of the photos is unknown, but the projection is known, which for real photos is not the case and it needs to be estimated separately, which also consumes a lot of data.
10:39 < TMA> (8 from 8 good angles, when the photos are near to each other the number goes up drastically)
10:42 < fleeky> for being lazy i quickly tried hyper3d.ai
10:42 < fleeky> i think that one took too many liberties , so i am going to try maybe running neuralrecon locally now
10:43 < fleeky> i want to do a lot of mask work + 3d printing so i need a good face mesh that is accurate .. 
10:45 < fleeky> also another dumb thing is it applied symetry to the face bah
10:56 < fleeky> i dont have windows but searching around i also found this : https://www.jawset.com/
10:56 < fleeky> theres also https://alicevision.org/
10:56 < fleeky> but meshroom / alicevision is a bit older
11:03  * L29Ah couldn't make meshroom work
11:06 < fleeky> yeah i kept getting camera errors also 
11:12 < hprmbridge> nmz787> I also failed to get meshroom to work with SEM images (was a relatively simple etched silicon trench that I was trying to get a depth estimate from)
11:12 < hprmbridge> nmz787> but some train locomotive engine dataset I found on instructables for meshroom worked fine
11:13 < hprmbridge> nmz787> (many many more images though, and much more complex scene... which presumably helped vs my quite smooth images)
11:45 -!- RangerMauve [m-4bpbmo@matrix.mauve.moe] has quit [Remote host closed the connection]
11:48 -!- RangerMauve [m-4bpbmo@matrix.mauve.moe] has joined #hplusroadmap
12:33 -!- okelano [~okelano@m90-131-41-219.cust.tele2.lt] has joined #hplusroadmap
13:56 -!- okelano [~okelano@m90-131-41-219.cust.tele2.lt] has quit [Quit: Client closed]
14:04 -!- okelano [~okelano@m90-131-41-219.cust.tele2.lt] has joined #hplusroadmap
15:24 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
15:27 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 252 seconds]
15:27 -!- stipa_ is now known as stipa
16:30 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 260 seconds]
16:49 < fleeky> omg yusss
16:49 < fleeky> i actually found my revopoint 3d scanner
16:50 < fleeky> TMA: long story short 3d scanner is far superior :( ai stuff is kind of a joke for any amount of accuracy , still might try the slam methods at some point but now that i've found my scanner i have to move on with my project
16:50 < fleeky> the scanner has 0.13mm accuracy
16:50 < fleeky> the biggest issue is that revopoint only has windows software and the software is terrible , so i gotta try to get it to run under wine bah
17:00 < fleeky> https://7rnx.net/tmp/mask.jpg
17:00 < fleeky> guess which is the 3d scanner and which one was using hyperfy.ai lol
19:51 < hprmbridge> nmz787> I think for my playground idea, the smoother one might be more idealized, vs all the splinters in the boards now, etc
19:52 < hprmbridge> nmz787> But i suppose that's a smooth filter operation away, with the better data
20:10 < fleeky> nmz787: if you dont need accuracy the smoother one is also probably easier to clean up
20:12 < fleeky> https://7rnx.net/tmp/mask1.jpg
20:13 < fleeky> phew 
20:13 < fleeky> that was a lot of clean up
20:13 < ike8> broken url
20:14 < fleeky> one thing that the scanner did not pick up was my hair so i need to cover my head next time with a cloth to be able to pick up my skull shape
20:14 < fleeky> ike8: reload
20:14 < ike8> fixed
20:15 < fleeky> but even still i can start to build masks off of this and print them out , i only pray that the scanner got the scale right haha
20:15 < fleeky> hopefully it will not be too much more before mask time though
20:16 < fleeky> the actual mesh data is atrocious 
20:16 < fleeky> like there are skin "layers" overlapping eachother
20:16 < fleeky> so basic decimation / clean up techniques don't work , you have to retop
20:16 < fleeky> but then also the retopo is hard because of the millimeter off scan slice differences
20:25 -!- flyback [~flyback@2601:540:c700:2380:443c:a9e8:cff5:85dd] has quit [Quit: Leaving]
20:25 -!- flyback [~flyback@2601:540:c700:2380:b811:75fa:25d:bdff] has joined #hplusroadmap
22:07 < fenn> https://dirtypcbs.com/about some kinda slowboat direct from china PCB and other stuffs manufacturing service related to hackaday
--- Log closed Sun May 25 00:00:53 2025