I´m working with contour plugin and a video camera input at 60fps, and i want to proyect only the original video input not the contour node video out (runs only at 25fps).
It is possible to “bypass” the contour node video output or something like that?
if you’re usig the latest addonpack the Contour has a new pin called Show Filtered which when set to 0 allows you to get teh unchanged video from the input piped through for display.
however, if alone enabling the contournode brings you down to 25fps and you need it enabled you won’t be able to view 60fps of videoinput on the same machine.
not sure if i got you right though…so does that help?
i don’t really get what you’re up to…
your camera input is 60fps. ok. you pipe it through contour (which takes a lot of power) so your patch runs only at 25fps anymore. so while contouring is running at 25fps you obviously can’t watch your video at 60fps at the same time, right?
Hi Joreg,
I´ll try to explain that with my poor English ;)
I think the problem is the limitation getting the video from the Freeframe plugin node, it´s not about how much power takes the plugin, my machine is not bad and i have enough power.
Yesterday i could render in the same window a video direct from the video texture and the data (video and contour data) from the Freeframecontour node but only if i worked only with one render window in the laptop screen, when i´ve tried to move the render window to the second screen, the contour plugin crashed and i have not video signal anymore.
This crash is related in the link of my last post (i think).
Originaly my patch runs at 60 fps but with the inconvenience that i have to use just one video signal (the plugin video output) that runs only at 25fps.
You can see here more or less that the patch is running very fluid with fluidsolver at 60fps :) Link
Maybe it´s a lot for a freeframe plugin but, it´s posible to compile it without the vieo output? it´s really needed?. allways you can catch the VideoIn transforming it into videotexture (much better).
I have not so much idea about programming, sorry if i say something stupid.
Hi Joreg,
I´ll try to explain that with my poor English ;)
I think the problem is the limitation getting the video from the Freeframe plugin node, it´s not about how much power takes the plugin, my machine is not bad and i have enough power.
Yesterday i could render in the same window a video direct from the video texture and the data (video and contour data) from the Freeframecontour node but only if i worked only with one render window in the laptop screen, when i´ve tried to move the render window to the second screen, the contour plugin crashed and i have not video signal anymore.
This crash is related in the link of my last post (i think).
Originaly my patch runs at 60 fps but with the inconvenience that i have to use just one video signal (the plugin video output) that runs only at 25fps.
You can see here more or less that the patch is running very fluid with fluidsolver at 60fps :) Link
Maybe it´s a lot for a freeframe plugin but, it´s posible to compile it without the vieo output? it´s really needed?. allways you can catch the VideoIn transforming it into videotexture (much better).
I have not so much idea about programming, sorry if i say something stupid.
I think the problem is the limitation getting the video from the Freeframe plugin node,
so which limitation do you see there? the contour node has a pin called Show Filtered: if enabled you get the thresholded video on contours output. if disabled you see the unchanged incoming video, right?
Yesterday i could render in the same window a video direct from the video texture and the data (video and contour data) from the Freeframecontour node but only if i worked only with one render window in the laptop screen, when i´ve tried to move the render window to the second screen, the contour plugin crashed and i have not video signal anymore.
ok, if the problem is that you want to view the same video on 2 different screens (in dualview mode), then right, that is not possible.
Originaly my patch runs at 60 fps but with the inconvenience that i have to use just one video signal (the plugin video output) that runs only at 25fps.
Maybe it´s a lot for a freeframe plugin but, it´s posible to compile it without the vieo output? it´s really needed?. allways you can catch the VideoIn transforming it into videotexture (much better).
those two paragraphs still make me think that i don’t exactly get your problem.
from your initial post:
“It is possible to “bypass” the contour node video output or something like that?”
how you mean bypass?! is it that you don’t want to see the video-output? you want contour to compute the contours, but don’t want to see the video at all? then put the videotexture on a quad and disable the quad. contour has to be connected to a renderer downstream to run at all. and it is not the videooutput that makes contour slow, but the computation of the contours itself. so only disabling the video-output won’t save much performance.
are we getting closer or am i still completely off topic?
I think the problem is the limitation getting the video from the Freeframe plugin node,
so which limitation do you see there? the contour node has a pin called Show Filtered: if enabled you get the thresholded video on contours output. if disabled you see the unchanged incoming video, right?
Yesterday i could render in the same window a video direct from the video texture and the data (video and contour data) from the Freeframecontour node but only if i worked only with one render window in the laptop screen, when i´ve tried to move the render window to the second screen, the contour plugin crashed and i have not video signal anymore.
ok, if the problem is that you want to view the same video on 2 different screens (in dualview mode), then right, that is not possible.
Originaly my patch runs at 60 fps but with the inconvenience that i have to use just one video signal (the plugin video output) that runs only at 25fps.
Maybe it´s a lot for a freeframe plugin but, it´s posible to compile it without the vieo output? it´s really needed?. allways you can catch the VideoIn transforming it into videotexture (much better).
those two paragraphs still make me think that i don’t exactly get your problem.
from your initial post:
“It is possible to “bypass” the contour node video output or something like that?”
how you mean bypass?! is it that you don’t want to see the video-output? you want contour to compute the contours, but don’t want to see the video at all? then put the videotexture on a quad and disable the quad. contour has to be connected to a renderer downstream to run at all. and it is not the videooutput that makes contour slow, but the computation of the contours itself. so only disabling the video-output won’t save much performance.
are we getting closer or am i still completely off topic?
Yes you are on topic, but maybe i have problems to explain this in English correctly.
when i use the video output of the contour node, it only runs at 25fps even if i disable the show filtered pin, and i say again that it can maybe be a limitation of the freeframe plugins.
I´ve used a lot of freeframe plugins with my 60fps camera and lot of them (maybe all) have this problem, it only runs at 25 fps.
Maybe the ground of the problem is the asvideo node?
I´ve switched the sync pin to “none” and i have better performance but not the same when i don´t use contour plugin.
A question:
I cant´t understand why the contour node needs a lot of power but in the other side i can render my original video signal at 60fps in the same window that the contour node video output, and a lot of grafic stuff at 60fps with the data of the contour node (at 25fps).
All that i want to say with all those things is, the plugin doesn´t make my computer slower.
in my opinion that smells like a programming problem with the freeframe plugins or something like that.
Have i to remember that i have no idea about programming ;) ??
Thanks a lot and i hope you can understand my point of view.
If you are in berlin we can meet and talk about that in front of a computer, (and a bier :)
Yes you are on topic, but maybe i have problems to explain this in English correctly.
when i use the video output of the contour node, it only runs at 25fps even if i disable the show filtered pin, and i say again that it can maybe be a limitation of the freeframe plugins.
I´ve used a lot of freeframe plugins with my 60fps camera and lot of them (maybe all) have this problem, it only runs at 25 fps.
Maybe the ground of the problem is the asvideo node?
I´ve switched the sync pin to “none” and i have better performance but not the same when i don´t use contour plugin.
A question:
I cant´t understand why the contour node needs a lot of power but in the other side i can render my original video signal at 60fps in the same window that the contour node video output, and a lot of grafic stuff at 60fps with the data of the contour node (at 25fps).
All that i want to say with all those things is, the plugin doesn´t make my computer slower.
in my opinion that smells like a programming problem with the freeframe plugins or something like that.
Have i to remember that i have no idea about programming ;) ??
Thanks a lot and i hope you can understand my point of view.
If you are in berlin we can meet and talk about that in front of a computer, (and a bier :)
when i use the video output of the contour node, it only runs at 25fps even if i disable the show filtered pin, and i say again that it can maybe be a limitation of the freeframe plugins.
ok, i would probably not call it a “limitation” but just a fact, that the contouring algorithms are quite demanding and will certainly cause vvvv to slow down. of course there is always the chance that optimizing contours source could improve performance (remember: it is opensource).
Maybe the ground of the problem is the asvideo node?
aha!? you haven’t mentioned using that one before.
I´ve switched the sync pin to “none” and i have better performance but not the same when i don´t use contour plugin.
true: in beta=22 the default for the Reference Clock pin is “None” meaning the output will be as fast as possible. so if you notice higher framerates with the reference clock set to “none” that is all that you will get.
I cant´t understand why the contour node needs a lot of power but in the other side i can render my original video signal at 60fps in the same window that the contour node video output, and a lot of grafic stuff at 60fps with the data of the contour node (at 25fps).
again, i don’t get what you’re saying here. sorry. as mentioned above: contour consumes power, as long as its Enabled pin is set to 1. if you disable it you should see higher fps, but no contour-data output of course.
I think more or less my questions are resolved, contour node is very desirability, but for performances that really needs very fast response (Dance for example) maybe is better try other ways ,like pipet and working with luminance data or something like that.
please check debug node how CPU power is taken.
Sometimes it helps if you scale down your video before tunneling it into asvideo.node. By a factor of 8,4,2 and result of contour might be still good. See patch
Just a side node, the asvideo node makes more or less everything very slow and similarly with the videotexture nodes although not quite as much.
you can scale your video either by entering the desired dimensions in the videoin node or using Scale (DShow9) if you do this in place of the asvideo solution you should get much better performan ce.
contour and its friend trautner are both freeframe nodes and operates on the gpu.
Asvideo and videotexture moves the video data from and to the gpu. this process involves copying a lot of data back and forth between cpu and gpu, this is somewhat slow.
I have come to the conclusion that I need to avoid this as much as possible.
I´m trying to avoid to use Asvideo, but i need to transform the video signal with a few effects before Freeframe Contour node (BG substraction, Filter Noise, 2 x Gaussian Blur) to keep a good and stable tracking, and this is the only way that i can imagine using only one freeframe plugin (contour node).
Have you noticed the same performance using a freeframe chain effects (BG substraction, etc)?
or do you think that it´s better to make the BG Sus, etc with shaders and then convert the video texture to asvideo before the Freeframe Contour node?
Thanks a lot for the information about your experience,