What Rob suggested should work. Another approach, if you do not want to create a low res version will be to set up a custom attribute on point of our interest. You can visually select these points and run a script to set the attribute value, then in your ICE graph you can query filter the point based on the custom attribute. Although it would not be presumably as fast as Rob's approach.
On Wed, Jun 6, 2012 at 6:58 AM, Ciaran Moloney <[email protected]>wrote: > Any chance you have a dependency cycle between the low and high res > clouds? You usually have to cache out the low res cloud to avoid such > things... > > Ciaran > > > On Wed, Jun 6, 2012 at 11:26 AM, Sebastian Kowalski <[email protected]>wrote: > >> Hey Rob, >> >> thanks for the good suggestion, i will try how far i can get away with a >> lower res this time. >> Besides that, i am running in massive bugs when working with different >> clouds, reading id's, jumping contextes. >> >> querying closest point from a cloud to another (just to make sure, i am >> using only positions i really want) giving me soo much pain when it comes >> down to change context (per point array to object array, and then writing >> it back cloud). >> values change randomly, making no sense at all.. first time i am really >> pissed working in ice. >> i will see if i can provide a repro scene. >> if i calm down >> >> >> >> Am 06.06.2012 um 11:35 schrieb Rob Chapman: >> >> >> Hi Sebastian, >>> >>> what are you trying to do with the 10 million+ point cloud? closest >>> location to itself is going to be painful yes :) when the point >>> clouds are reaching that high sometimes its best to have a similar >>> pointcloud around but with far fewer points. like say 1%. >>> Surprisingly it is a heck of a lot faster doing closest points lookup >>> from cloud A(10mill+) to a lowres cloud B (100k+) than to itself. >>> >>> So workflow is:- >>> >>> take original cloud A at 10mil+ duplicate and reduce down to Cloud B >>> 1% with Ice nodes 'test random probability' and 'delete point' >>> remember to freeze this cloud so the delete point is not live >>> >>> do all closest location lookup and additional processing on Cloud B >>> >>> when finished do closest location lookup cloud A to B to transfer >>> results back to A >>> >>> >>> Its not perfect solution for everything but in some cases with eg >>> lighting pointclouds or working out densities in ICE it beats waiting >>> around for those 10mill+ points to work themselves out! >>> >>> you may be able to do something cleverer with only the one high res >>> pointcloud but for me this level of abstraction works out ok >>> >>> cheers >>> >>> Rob >>> >>> >>> >>> >>> >>> On 6 June 2012 09:26, Sebastian Kowalski <[email protected]> wrote: >>> >>>> hey list, >>>> >>>> i dont need to check all pointpositions for closest points, just a few. >>>> but >>>> i cant figure out how to filter them out. >>>> got this huge cloud (10.000.000+ points) and softimage processes >>>> neighbours >>>> pts forever... >>>> >>>> the closest point node wont accept arrays, and i am afraid we cant get a >>>> filtered set >>>> >>>> any ideas? >>>> >>>> thanks in advance >>>> >>>> >>>> ps. ive tried to just calculate distance between all points with just >>>> array >>>> resizing, this is fast, but i cant get near a million without >>>> skyrocketing >>>> memory consumption, and therefore system freeze ups. >>>> >>>> >>>> >> > --
