Re: Re: [Nuke-users] DeepToPoints bug?
Hey guys, Thanks for your help - I've heard of the BBox issue on deep files so I'll check that out if I can reproduce the bug again. If I still see the bug in 6.3v9 and 7 then I'll submit it to the Foundry. Michael On 12 February 2013 02:53, Patrick Heinen wrote: > Yes, sorry Ivan you're absolutely right! It won't influence it in such a > heavy way. My mistake. > Michael, I just tried reproducing the bug you mentioned over here with a > couple of vrst files. But mine look all right. However I can reliably > reproduce the symptoms you mention when using a DeepCrop before the > DeepToPoints, that alters the bounding box. > cheers > Patrick > > - Original Message - > From: ivanbusqu...@gmail.com > To: nuke-users@support.thefoundry.co.uk > Date: 12.02.2013 02:47:13 > Subject: Re: [Nuke-users] DeepToPoints bug? > > > > Hi Michael, > > > > Hope all is going well. Can't verify this as I'm not in front of Nuke, > but > > that does sound like a bug to me (although I've never observed such > > behaviour) > > > > Patrick, even if the near and far clipping planes contribute to the > > projection matrix, they would not alter it in such a way that projecting > a > > point on to a given depth would give different results. > > The direction of the output vector should be the same (barring minor > > rounding errors), so the point of that vector that lies at a certain > depth > > from camera will also remain the same. > > > > > > > > > > On Mon, Feb 11, 2013 at 4:46 PM, Patrick Heinen < > > mailingli...@patrickheinen.com> wrote: > > > >> Hey Michael, > >> > >> a bug I know of in combination with the DeepToPoints is that if the > >> bounding box is not equal to the format, that can cause weird behaviour, > >> similar to the stretching you mention. That the near clipping plane > changes > >> the position in 3d space seems to actually be no bug, but a normal > >> behaviour. The clipping planes influence the camera projection matrix, > and > >> thus it is normal getting different result for the DeepToPoints, as it > >> multiplies your Vector with the inverse of the projection matrix to get > the > >> position in world space. > >> > >> So export your camera from your 3d application or use the information > >> rendered to the metadata to build your cam and don't change the > settings of > >> it. > >> Hope that helps, if you need it more detailed I can explain it further > >> tomorrow. > >> I'm actually using vrst files aswell ;) > >> > >> cheers > >> Patrick > >> > >> Am 11.02.2013 um 20:07 schrieb Michael Garrett: > >> > >> > Hey, > >> > > >> > I've found on Nuke 6.3v8 on Windows that the near clipping plane of a > >> Camera plugged into DeepToPoints will affect how the point cloud is > cast. > >> The depth of samples gets thrown off, ie, they go non-linear and stretch > >> out, when the near clipping plane is reduced, as if you've done a gamma > >> curve on the deep data. > >> > > >> > I want to try this in 6.3v9 and 7.x as soon as I can, to see if it's a > >> bug that's been fixed. We're using .vrst files in this case, not that it > >> should make any difference. > >> > > >> > Typically I just re-checked the scene and it's working fine now... > >> > > >> > Has anyone else experienced this? > >> > > >> > Thanks, > >> > Michael > >> > ___ > >> > Nuke-users mailing list > >> > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > >> > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > >> > >> ___ > >> Nuke-users mailing list > >> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > >> > > ___ > > Nuke-users mailing list > > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > ___ > Nuke-users mailing list > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: Re: [Nuke-users] DeepToPoints bug?
Yes, sorry Ivan you're absolutely right! It won't influence it in such a heavy way. My mistake. Michael, I just tried reproducing the bug you mentioned over here with a couple of vrst files. But mine look all right. However I can reliably reproduce the symptoms you mention when using a DeepCrop before the DeepToPoints, that alters the bounding box. cheers Patrick - Original Message - From: ivanbusqu...@gmail.com To: nuke-users@support.thefoundry.co.uk Date: 12.02.2013 02:47:13 Subject: Re: [Nuke-users] DeepToPoints bug? > Hi Michael, > > Hope all is going well. Can't verify this as I'm not in front of Nuke, but > that does sound like a bug to me (although I've never observed such > behaviour) > > Patrick, even if the near and far clipping planes contribute to the > projection matrix, they would not alter it in such a way that projecting a > point on to a given depth would give different results. > The direction of the output vector should be the same (barring minor > rounding errors), so the point of that vector that lies at a certain depth > from camera will also remain the same. > > > > > On Mon, Feb 11, 2013 at 4:46 PM, Patrick Heinen < > mailingli...@patrickheinen.com> wrote: > >> Hey Michael, >> >> a bug I know of in combination with the DeepToPoints is that if the >> bounding box is not equal to the format, that can cause weird behaviour, >> similar to the stretching you mention. That the near clipping plane changes >> the position in 3d space seems to actually be no bug, but a normal >> behaviour. The clipping planes influence the camera projection matrix, and >> thus it is normal getting different result for the DeepToPoints, as it >> multiplies your Vector with the inverse of the projection matrix to get the >> position in world space. >> >> So export your camera from your 3d application or use the information >> rendered to the metadata to build your cam and don't change the settings of >> it. >> Hope that helps, if you need it more detailed I can explain it further >> tomorrow. >> I'm actually using vrst files aswell ;) >> >> cheers >> Patrick >> >> Am 11.02.2013 um 20:07 schrieb Michael Garrett: >> >> > Hey, >> > >> > I've found on Nuke 6.3v8 on Windows that the near clipping plane of a >> Camera plugged into DeepToPoints will affect how the point cloud is cast. >> The depth of samples gets thrown off, ie, they go non-linear and stretch >> out, when the near clipping plane is reduced, as if you've done a gamma >> curve on the deep data. >> > >> > I want to try this in 6.3v9 and 7.x as soon as I can, to see if it's a >> bug that's been fixed. We're using .vrst files in this case, not that it >> should make any difference. >> > >> > Typically I just re-checked the scene and it's working fine now... >> > >> > Has anyone else experienced this? >> > >> > Thanks, >> > Michael >> > ___ >> > Nuke-users mailing list >> > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ >> > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> >> ___ >> Nuke-users mailing list >> Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users >> > ___ > Nuke-users mailing list > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] DeepToPoints bug?
Hi Michael, Hope all is going well. Can't verify this as I'm not in front of Nuke, but that does sound like a bug to me (although I've never observed such behaviour) Patrick, even if the near and far clipping planes contribute to the projection matrix, they would not alter it in such a way that projecting a point on to a given depth would give different results. The direction of the output vector should be the same (barring minor rounding errors), so the point of that vector that lies at a certain depth from camera will also remain the same. On Mon, Feb 11, 2013 at 4:46 PM, Patrick Heinen < mailingli...@patrickheinen.com> wrote: > Hey Michael, > > a bug I know of in combination with the DeepToPoints is that if the > bounding box is not equal to the format, that can cause weird behaviour, > similar to the stretching you mention. That the near clipping plane changes > the position in 3d space seems to actually be no bug, but a normal > behaviour. The clipping planes influence the camera projection matrix, and > thus it is normal getting different result for the DeepToPoints, as it > multiplies your Vector with the inverse of the projection matrix to get the > position in world space. > > So export your camera from your 3d application or use the information > rendered to the metadata to build your cam and don't change the settings of > it. > Hope that helps, if you need it more detailed I can explain it further > tomorrow. > I'm actually using vrst files aswell ;) > > cheers > Patrick > > Am 11.02.2013 um 20:07 schrieb Michael Garrett: > > > Hey, > > > > I've found on Nuke 6.3v8 on Windows that the near clipping plane of a > Camera plugged into DeepToPoints will affect how the point cloud is cast. > The depth of samples gets thrown off, ie, they go non-linear and stretch > out, when the near clipping plane is reduced, as if you've done a gamma > curve on the deep data. > > > > I want to try this in 6.3v9 and 7.x as soon as I can, to see if it's a > bug that's been fixed. We're using .vrst files in this case, not that it > should make any difference. > > > > Typically I just re-checked the scene and it's working fine now... > > > > Has anyone else experienced this? > > > > Thanks, > > Michael > > ___ > > Nuke-users mailing list > > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > > ___ > Nuke-users mailing list > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users > ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
Re: [Nuke-users] DeepToPoints bug?
Hey Michael, a bug I know of in combination with the DeepToPoints is that if the bounding box is not equal to the format, that can cause weird behaviour, similar to the stretching you mention. That the near clipping plane changes the position in 3d space seems to actually be no bug, but a normal behaviour. The clipping planes influence the camera projection matrix, and thus it is normal getting different result for the DeepToPoints, as it multiplies your Vector with the inverse of the projection matrix to get the position in world space. So export your camera from your 3d application or use the information rendered to the metadata to build your cam and don't change the settings of it. Hope that helps, if you need it more detailed I can explain it further tomorrow. I'm actually using vrst files aswell ;) cheers Patrick Am 11.02.2013 um 20:07 schrieb Michael Garrett: > Hey, > > I've found on Nuke 6.3v8 on Windows that the near clipping plane of a Camera > plugged into DeepToPoints will affect how the point cloud is cast. The depth > of samples gets thrown off, ie, they go non-linear and stretch out, when the > near clipping plane is reduced, as if you've done a gamma curve on the deep > data. > > I want to try this in 6.3v9 and 7.x as soon as I can, to see if it's a bug > that's been fixed. We're using .vrst files in this case, not that it should > make any difference. > > Typically I just re-checked the scene and it's working fine now... > > Has anyone else experienced this? > > Thanks, > Michael > ___ > Nuke-users mailing list > Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
[Nuke-users] DeepToPoints bug?
Hey, I've found on Nuke 6.3v8 on Windows that the near clipping plane of a Camera plugged into DeepToPoints will affect how the point cloud is cast. The depth of samples gets thrown off, ie, they go non-linear and stretch out, when the near clipping plane is reduced, as if you've done a gamma curve on the deep data. I want to try this in 6.3v9 and 7.x as soon as I can, to see if it's a bug that's been fixed. We're using .vrst files in this case, not that it should make any difference. Typically I just re-checked the scene and it's working fine now... Has anyone else experienced this? Thanks, Michael ___ Nuke-users mailing list Nuke-users@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/ http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
