Re: Test inside geometry failing
Cool! I always wanted to add the ray intersect test to make this thing more robust, but have long since run out of energy... Ciaran On Wed, May 8, 2013 at 1:14 AM, Eric Thivierge ethivie...@hybride.comwrote: Thanks guys, Ciaran, your addon actually works exceptionally well in my case. The slider for the Override_test_distance gets rid of the errant areas very very well. Not to mention generating this particle cloud is super fast now. You've saved me some late night work today. :) I did try to roll my own but was still getting the unwanted areas. The tips on how to build it are appreciated as it may help in future cases. Cheers, Eric Thivierge === Character TD / RnD Hybride Technologies On 07/05/2013 7:23 PM, Ciaran Moloney wrote: IIRC implementing the ray intersect technique is problematic in ICE, since the raycast node only returns the first intersection. You'd have to manually shoot subsequent rays after every hit. On Wed, May 8, 2013 at 12:06 AM, Raffaele Fragapane raffsxsil...@googlemail.com wrote: You could write/wire the test yourself. Point inside hull is fairly trivial outside of fringe cases, shoot a ray in any direction, if it intersects an odd number of times, it's outside, even, it's inside. You might want multiple rays and take the median of the result, or work conservatively and flag it even if just one tests true. You can further refine this with a force field approach, run a distance from hull (closest location on geo), and if the point is close enough to a location with a normal facing away from its distance vector remove it, and remove anything within a certain threshold (very close) indiscriminately. Testing against denser geometry also helps, even if it's more expensive, sparse and variedly populated topology tends to introduce more room for error. Lastly, a triangulated mesh, especially if you have a certain degree of dis-planarity in your geo on large enough areas, will also improve things in my experience. On Wed, May 8, 2013 at 4:54 AM, Eric Thivierge ethivie...@hybride.comwrote: Anyone notice how many times this just doesn't work correctly? Even toggling the Closed volume inside often doesn't give the correct results as well... I have a voxel setup with particles and testing inside a geometry to keep the ones that are within, works, however there are many points still outside the geo that remain and aren't deleted. The ones left outside changes each frame as well. Anyone have any sure fire methods / workarounds that don't involve having helper nulls to delete the remaining ones outside? -- Eric Thivierge === Character TD / RnD Hybride Technologies -- Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!
Re: Test inside geometry failing
Is this the thing where you get giant cubes of bad volume data, or something else? On 7 May 2013 19:54, Eric Thivierge ethivie...@hybride.com wrote: Anyone notice how many times this just doesn't work correctly? Even toggling the Closed volume inside often doesn't give the correct results as well... I have a voxel setup with particles and testing inside a geometry to keep the ones that are within, works, however there are many points still outside the geo that remain and aren't deleted. The ones left outside changes each frame as well. Anyone have any sure fire methods / workarounds that don't involve having helper nulls to delete the remaining ones outside? -- Eric Thivierge === Character TD / RnD Hybride Technologies
Re: Test inside geometry failing
maybe instead of just doing it in houdini he could use SDFs? http://blog.blackredking.org/?p=47 On Tue, May 7, 2013 at 12:34 PM, Peter Agg peter@googlemail.com wrote: Yeah, that's my only suggestion really - I've had quite a few problems before where I had very dense particles being culled by complex geo and on random frames you'd get very distinct cubes where the volume calculations have freaked out. Never found a particularly useful workaround, I'm afraid. Simplifying the geo has helped, sometimes not grouping them together and doing several tests on individual parts did. But sometimes that just shifts the problem to different frames - for the project I was on we had one of those unfortunate 'just do it in Houdini' moments. Good luck! :) On 7 May 2013 20:29, Eric Thivierge ethivie...@hybride.com wrote: Not necessarily cubes. If I run a smooth op on the geo it fixes much of it (although this is not a solution for me) so it's got to be the internal methods not liking something about the geometry. Eric Thivierge === Character TD / RnD Hybride Technologies On 07/05/2013 3:26 PM, Peter Agg wrote: Is this the thing where you get giant cubes of bad volume data, or something else? On 7 May 2013 19:54, Eric Thivierge ethivie...@hybride.com wrote: Anyone notice how many times this just doesn't work correctly? Even toggling the Closed volume inside often doesn't give the correct results as well... I have a voxel setup with particles and testing inside a geometry to keep the ones that are within, works, however there are many points still outside the geo that remain and aren't deleted. The ones left outside changes each frame as well. Anyone have any sure fire methods / workarounds that don't involve having helper nulls to delete the remaining ones outside? -- Eric Thivierge === Character TD / RnD Hybride Technologies
Re: Test inside geometry failing
You could write/wire the test yourself. Point inside hull is fairly trivial outside of fringe cases, shoot a ray in any direction, if it intersects an odd number of times, it's outside, even, it's inside. You might want multiple rays and take the median of the result, or work conservatively and flag it even if just one tests true. You can further refine this with a force field approach, run a distance from hull (closest location on geo), and if the point is close enough to a location with a normal facing away from its distance vector remove it, and remove anything within a certain threshold (very close) indiscriminately. Testing against denser geometry also helps, even if it's more expensive, sparse and variedly populated topology tends to introduce more room for error. Lastly, a triangulated mesh, especially if you have a certain degree of dis-planarity in your geo on large enough areas, will also improve things in my experience. On Wed, May 8, 2013 at 4:54 AM, Eric Thivierge ethivie...@hybride.comwrote: Anyone notice how many times this just doesn't work correctly? Even toggling the Closed volume inside often doesn't give the correct results as well... I have a voxel setup with particles and testing inside a geometry to keep the ones that are within, works, however there are many points still outside the geo that remain and aren't deleted. The ones left outside changes each frame as well. Anyone have any sure fire methods / workarounds that don't involve having helper nulls to delete the remaining ones outside? -- Eric Thivierge === Character TD / RnD Hybride Technologies -- Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!
Re: Test inside geometry failing
Thanks guys, Ciaran, your addon actually works exceptionally well in my case. The slider for the Override_test_distance gets rid of the errant areas very very well. Not to mention generating this particle cloud is super fast now. You've saved me some late night work today. :) I did try to roll my own but was still getting the unwanted areas. The tips on how to build it are appreciated as it may help in future cases. Cheers, Eric Thivierge === Character TD / RnD Hybride Technologies On 07/05/2013 7:23 PM, Ciaran Moloney wrote: IIRC implementing the ray intersect technique is problematic in ICE, since the raycast node only returns the first intersection. You'd have to manually shoot subsequent rays after every hit. On Wed, May 8, 2013 at 12:06 AM, Raffaele Fragapane raffsxsil...@googlemail.com mailto:raffsxsil...@googlemail.com wrote: You could write/wire the test yourself. Point inside hull is fairly trivial outside of fringe cases, shoot a ray in any direction, if it intersects an odd number of times, it's outside, even, it's inside. You might want multiple rays and take the median of the result, or work conservatively and flag it even if just one tests true. You can further refine this with a force field approach, run a distance from hull (closest location on geo), and if the point is close enough to a location with a normal facing away from its distance vector remove it, and remove anything within a certain threshold (very close) indiscriminately. Testing against denser geometry also helps, even if it's more expensive, sparse and variedly populated topology tends to introduce more room for error. Lastly, a triangulated mesh, especially if you have a certain degree of dis-planarity in your geo on large enough areas, will also improve things in my experience. On Wed, May 8, 2013 at 4:54 AM, Eric Thivierge ethivie...@hybride.com mailto:ethivie...@hybride.com wrote: Anyone notice how many times this just doesn't work correctly? Even toggling the Closed volume inside often doesn't give the correct results as well... I have a voxel setup with particles and testing inside a geometry to keep the ones that are within, works, however there are many points still outside the geo that remain and aren't deleted. The ones left outside changes each frame as well. Anyone have any sure fire methods / workarounds that don't involve having helper nulls to delete the remaining ones outside? -- Eric Thivierge === Character TD / RnD Hybride Technologies -- Our users will know fear and cower before our software! Ship it! Ship it and let them flee like the dogs they are!
