00:00.54 | *** join/#neo900 XDS2010____ (sid1218@gateway/web/irccloud.com/x-jlnnaskdadxuvgmb) |
00:03.02 | *** join/#neo900 ZeroChaos (~zerochaos@gentoo/developer/pentoo/zerochaos) |
00:04.51 | DocScrutinizer05 | for that approach it's probably sufficient for all but the most potato shaped objects to find the "tangent" plane that touches bottom (the largest surface) in 3 points, then have a parallel plane that touches top in one point. Then do same for the 4 sides that you did for top |
00:05.45 | DocScrutinizer05 | actually for bottom&top, to rotate the cubicle around Z axis accordingly |
00:06.08 | *** join/#neo900 FredrIQ (~FIQ@fiq.se) |
00:06.42 | DocScrutinizer05 | this approach however will _not_ (always) result in min(x*y*z) |
00:13.04 | DocScrutinizer05 | e.g when you got a sheet (parallel bottom and top) with irregular hexagon shape, the sides of enclosing cuboid could be defined by two sides of the hexagon that result in a suboptimal solution for x*y |
00:18.26 | wpwrak | yes, there are lots of troublesome shaped for pretty much anything. the algorithm would basically have to "understand" the design, which is kinda impossible. (well, i'd be scared if it suddenly did :) |
00:20.35 | wpwrak | but for things that are basically box-shaped, it may work. if the sides are too curved, then things get trickier. then one could try to fit tangent plane, have some distance metric that peaks for some not too far-off distance, then drops (for points that aren't part of that plane anyway), but that sounds complicated |
01:15.01 | *** join/#neo900 norly (~norly@enpas.org) |
01:59.27 | DocScrutinizer05 | however there are nifty sophisticated concepts that even place arbitrarily shaped forms on a large sheet (of steel) in a way so you get the minimum of waste cut away (called "Schachtler", working in Siemens Geraetewerk Erlangen to automatically cut case parts out of steel sheets), so I guess when even this can get solved algorithmically then your simple "find smallest volume enclosing cuboid" should also be feasible |
02:01.08 | DocScrutinizer05 | worst case with an iterative or evolutionary process |
02:14.34 | *** join/#neo900 Oksana (~chatzilla@129.94.239.199) |
02:28.06 | wpwrak | some of these things may also have a considerable amount of research behind them :) anyway, i was just looking for something simple. see if we can do better than manual alignment |
02:29.20 | wpwrak | that would be plan B: have virtual planes parellel to { xy, xz, yx } intersect the mesh, show the result, and rotate the mesh accordingly. |
03:18.15 | DocScrutinizer05 | cheerfully looks forward to GDC's quotation for proto o/ |
03:23.11 | Oksana | Cheerfully? |
04:25.04 | *** join/#neo900 chainsawbike (~chainsawb@unaffiliated/chainsawbike) |
05:24.05 | *** join/#neo900 M13 (~MirandaLS@83.149.35.9) |
06:17.07 | *** join/#neo900 Pali (~pali@Maemo/community/contributor/Pali) |
08:00.26 | *** join/#neo900 Kabouik (~quassel@147.99.218.243) |
10:10.22 | *** join/#neo900 kolp (~quassel@55d45cb4.access.ecotel.net) |
10:28.37 | wpwrak | sometimes, a bit of sleep helps :) just tought of this: pick a random sample of sets of pairs of points, calculate the angles between the imaginary line between them and any of the three planes in the coordinate system. plot the distribution. it should peak around multiples of 90 deg. then filter that and calculate better fit. |
10:28.49 | wpwrak | let's try this ... |
10:30.26 | wpwrak | btw, the reason why i'm interested in this problem is that i already know that slight tilts in the coordinate system come back to haunt you. so if i can clean up the raw data, that would make things much easier to work with later on.) |
10:31.32 | *** join/#neo900 Pali (~pali@Maemo/community/contributor/Pali) |
11:01.47 | *** join/#neo900 che1 (~che@p5B072937.dip0.t-ipconnect.de) |
11:36.45 | *** join/#neo900 mvaenskae (~mvaenskae@unaffiliated/mvaenskae) |
11:55.16 | DocScrutinizer05 | good |
13:44.25 | *** join/#neo900 gindating (~gindating@ec2-54-201-199-156.us-west-2.compute.amazonaws.com) |
13:49.25 | *** join/#neo900 silviof (~silviof@unaffiliated/silviof) |
14:13.28 | *** join/#neo900 nicksydney_ (~quassel@197.139-50-210.dynamic.dsl.syd.iprimus.net.au) |
14:19.39 | *** join/#neo900 Guest56347 (~FIQ@prgmr/customer/fiq) |
14:28.19 | *** join/#neo900 nicksydney (~quassel@197.139-50-210.dynamic.dsl.syd.iprimus.net.au) |
14:50.09 | drathir | mornin... |
16:06.27 | *** join/#neo900 sixwheeledbeast (~paul@host-92-29-203-193.as13285.net) |
16:49.32 | *** join/#neo900 bleb (~caleb@209-6-88-180.c3-0.smr-ubr1.sbo-smr.ma.cable.rcn.com) |
18:17.21 | *** join/#neo900 xes (~xes@unaffiliated/xes) |
18:30.23 | *** join/#neo900 mvaenskae (~mvaenskae@unaffiliated/mvaenskae) |
18:37.34 | josch_ | wpwrak: a while ago you wanted an algorithm to rotate a point cloud such that it fits into a rectangular cuboid ("3d rectangle") with the smallest volume |
18:38.08 | josch_ | wpwrak: this problem seems to be the same as finding an oriented minimum bounding box for a given point cloud |
18:38.36 | josch_ | wpwrak: there is lots of material of how to find the size and orientation of this minimum oriented bounding box - why not use that? |
18:50.55 | *** join/#neo900 M13 (~MirandaLS@2a02:17d0:d5:2300:d491:cea1:93fa:3491) |
19:46.49 | *** join/#neo900 Kabouik (~quassel@208.237.138.88.rev.sfr.net) |
21:00.50 | *** join/#neo900 bangsnap (~bangsnap@ec2-54-201-199-156.us-west-2.compute.amazonaws.com) |
21:27.19 | mvaenskae | i just dedusted my n900 and am surprised by how small it feels |
21:29.42 | *** join/#neo900 norly (~norly@enpas.org) |
22:53.57 | *** join/#neo900 mvaenskae (~mvaenskae@unaffiliated/mvaenskae) |
23:38.01 | DocScrutinizer05 | Oksana: received your mail. Will split into 3 units. Many thanks! :-) |
23:38.43 | Oksana | DocScrutinizer : Yep. Two donations, three units. |
23:38.49 | DocScrutinizer05 | :-D |
23:39.07 | Oksana | Just do not want to pay more in bank fees than I have to. :) |
23:39.07 | DocScrutinizer05 | it's a pleasure |
23:40.07 | DocScrutinizer05 | just to make sure: 3 series devices incl 3 cases, zero protoypes |
23:42.25 | DocScrutinizer05 | Oksana: right? |
23:43.09 | Oksana | Yes. Three 'mass-production' devices. Including cases, frames, however it is called. |
23:43.21 | DocScrutinizer05 | *nod* |
23:44.05 | mvaenskae | ohhh, device orders are growing :) nice to see |