madVR - high quality video renderer (GPU assisted)

[RBG]

Can somebody tell me what algorithm is used in NNEDI3 upscaling or at least point out where I can read about it? I can't find any useful source.

[nevcairiel]

Originally Posted by RBG

Can somebody tell me what algorithm is used in NNEDI3 upscaling or at least point out where I can read about it? I can't find any useful source.

NNEDI3 is the name of the algorithm.

[RBG]

Originally Posted by nevcairiel

NNEDI3 is the name of the algorithm.

Ok, and where can I read about this algorithm?

[Guest]

Do a google search for edge directed interpolation (EDI). tritical implements it with a neural network (NN). Further discussion should be in a new thread or an existing related thread.

[leeperry]

Originally Posted by huhn

CUVID deint is the same like dxva deint the case you linked me where dxva deint was very bad but CUVID was totally fine

I remember something along the lines of "pixel adaptive deinterlacing" in nvidia's sales pitch...and IIRC when I compared 576i25@50p between YADIF & non-HQ CUVID deinterlacing, the latter looked quite a bit better so HQ might further help.

The whole point for me with CUVID is that it runs on the GPU(using the PureVideo routines so it's not hogging mVR's potential cycles) & it especially doesn't work behind Reclock's back...I can just pop a PAL or NTSC interlaced video DVD in, LAV does the 2X magic, mVR rolls refresh rates automatically et voilà: 50/59.94fps video upscaled via NNEDI, what a fantastic way to give DVD's a second youth

James Cameron said that he would prefer a higher fps rate over a higher resolution and I haven't run any real-world comparison but there's a good chance that I would pick NNEDI'ed 540p59.94/576p50 DVD's over 25/29.97 720p.

Originally Posted by neuron2

Do a google search for edge directed interpolation (EDI). tritical implements it with a neural network (NN). Further discussion should be in a new thread or an existing related thread.

Yeah, even tritical isn't too verbose regarding how NNEDI achieves those impressive results......but this link is kinda clear on what EDI really does and how stunning it looks

[G_M_C]

Originally Posted by leeperry

[...]we'll see about that as it's undeniable that my 7850 looks outstanding with mVR [...]

Hear! Hear!

As does mine! Especially so since I can use NNEdi chroma upscaling for most situations on my HD7850.

[annovif]

I've understood that now, nvidia has 20-30% more efficiency with NNEDI than Ati .

[turbojet]

Another advantage I found of new window mode is it properly detects video when film mode is forced as 2:2.

Would it be possible to detect between 25/29 and 50/59 native fps rather than toggling double framerate deinterlacing? A lot of interlaced footage is only 25/29, there isn't any advantage to outputting this at 50/59 fps is there?

[nevcairiel]

Originally Posted by turbojet

A lot of interlaced footage is only 25/29, there isn't any advantage to outputting this at 50/59 fps is there?

If its interlaced, then it needs to be output at twice its rate (ie one frame for each field), or you lose information.

[Anime Viewer]

Originally Posted by leeperry

I remember something along the lines of "pixel adaptive deinterlacing" in nvidia's sales pitch...and IIRC when I compared 576i25@50p between YADIF & non-HQ CUVID deinterlacing, the latter looked quite a bit better so HQ might further help.

The whole point for me with CUVID is that it runs on the GPU(using the PureVideo routines so it's not hogging mVR's potential cycles) & it especially doesn't work behind Reclock's back...I can just pop a PAL or NTSC interlaced video DVD in, LAV does the 2X magic, mVR rolls refresh rates automatically et voilà: 50/59.94fps video upscaled via NNEDI, what a fantastic way to give DVD's a second youth

The problem with Nvidia hardware acceleration is things like this:

https://esports.geforce.com/forums/topic/527075/geforce-drivers/hardware-acceleration-decoding-issue/1

and

https://esports.geforce.com/forums/topic/616833/geforce-drivers/hardware-acceleration-decoding-issue-new-/

[DragonQ]

Originally Posted by turbojet

Another advantage I found of new window mode is it properly detects video when film mode is forced as 2:2.

Would it be possible to detect between 25/29 and 50/59 native fps rather than toggling double framerate deinterlacing? A lot of interlaced footage is only 25/29, there isn't any advantage to outputting this at 50/59 fps is there?

Interlaced footage by its nature is going to be 50/60 fps. If it's PsF (progressively-segmented-frames) then it should be weaved to 25/30 fps instead. nVidia's "video mode" deinterlacing algorithm detects this situation pretty well IME and will apply weaving as appropriate.

AMD's vector adaptive deinterlacer seems to detect this well too. Makes you wonder why so many TVs (even high end ones) seem to have issues detecting 2:2 cadence.

[Ceremony]

@madshi, if you ever get into (h)UMA, just give me a call/pm for testing. I switched to a AMD APU (A10-7850K)

[nevcairiel]

Originally Posted by Anime Viewer

The problem with Nvidia hardware acceleration is things like this:

https://esports.geforce.com/forums/topic/527075/geforce-drivers/hardware-acceleration-decoding-issue/1

and

https://esports.geforce.com/forums/topic/616833/geforce-drivers/hardware-acceleration-decoding-issue-new-/

The first issue is fixed on Maxwell cards, I didn't test the second, but it might as well be.
Its a very rare occurrence really, I've never encountered it in any real video, only weird short sample clips people post here.

If you want to compare, AMDs hardware acceleration is anything but glamorous.

[turbojet]

Originally Posted by nevcairiel

If its interlaced, then it needs to be output at twice its rate (ie one frame for each field), or you lose information.

29 and 59 look identical to me (dithering so not exactly)[/URL].
EDIT: removed video link, it wasn't native 29.

Originally Posted by DragonQ

Interlaced footage by its nature is going to be 50/60 fps. If it's PsF (progressively-segmented-frames) then it should be weaved to 25/30 fps instead. nVidia's "video mode" deinterlacing algorithm detects this situation pretty well IME and will apply weaving as appropriate.

AMD's vector adaptive deinterlacer seems to detect this well too. Makes you wonder why so many TVs (even high end ones) seem to have issues detecting 2:2 cadence.

I can only speak for nvidia but yes it does handle it well but madvr is using much more gpu outputting to 59 fps.

[6233638]

Originally Posted by Anime Viewer

The problem with Nvidia hardware acceleration is things like this

This is an issue with hardware acceleration in general. Maybe not those specific samples, but you will be able to find ones which do the same thing on AMD/Intel as well, I'm sure.

But this is extremely rare if you are watching commercial content i.e. Blu-ray discs. I've only seen it happen twice out of probably thousands of discs at this point.

[Asmodian]

Originally Posted by leeperry

The whole point for me with CUVID is that it runs on the GPU(using the PureVideo routines so it's not hogging mVR's potential cycles) & it especially doesn't work behind Reclock's back...I can just pop a PAL or NTSC interlaced video DVD in, LAV does the 2X magic, mVR rolls refresh rates automatically et voilà: 50/59.94fps video upscaled via NNEDI, what a fantastic way to give DVD's a second youth

The deinterlacing is not done entirely by PureVideo, that handles the decoding but not all the computation required for deinterlacing. At least this was true for VP5 and it is probably the same for VP6. Remember the GT520 when VP5 first came out that used lower quality deinterlacing because it didn't have enough shaders for the higher quality deinterlacing?

Originally Posted by turbojet

29 and 59 look identical to me (dithering so not exactly) on a 29 fps source.

Screenshots to show the difference or lack thereof between 29 and 59 fps? Try watching a panning shot, someone running, or something with motion.

If you do not trust nevcairiel on this point, this is from 100fps.com interlaced video
"Because of this time intermix (1 frame=time1+time2) it is impossible to:

1) deinterlace a frame
AND 2) keep 25 frames/second
AND 3) keep the full quality (=all information of a picture).
Impossible. You will have to alter at least one of those points. Except, when there was no motion."

[bombadilio]

Anyone figure out how to get this to work with 4k video started using SVP and cant figure out the best way to upconvert 4k video to 60 fps

[Asmodian]

Originally Posted by bombadilio

Anyone figure out how to get this to work with 4k video started using SVP and cant figure out the best way to upconvert 4k video to 60 fps

What to work with 4k video? madVR works with 4k 60 fps video for me.

Originally Posted by 6233638

But this is extremely rare if you are watching commercial content i.e. Blu-ray discs. I've only seen it happen twice out of probably thousands of discs at this point.

It seems like it is usually a problem with broadcast streams more than blu-ray as it is due to errors or other issues with the stream that are unlikely to make it onto a pressed disc.

I think AMD has recently put some effort into making their hardware decoding more resilient and I have been unable to find any recent reports of issues like those but there was this example from 2010.

Of course they have their own odd random issues.

[turbojet]

Asmodian: I'm not talking about playing native 59 fps at 29 or 59, 59 would be optimal and that's what all those samples (except mousse_smooth_big.avi is kind of close, duplicate frames to go from 24 to 50) on 100fps is discussing. I'm talking about native 29 fps played at 59 fps by duplicating every frame. Is there some advantage to it or is it just a waste of gpu and would be better off just playing at 29 fps?

Nevermind the video sample above, it played in wtv as 29 and 59 fps but detects as 3:2 in mpg. I think it's a potplayer issue, resolution is detected wrong too.

Here's a native 29 fps video. Screenshots of a 29 fps frame is bit identical to the 2 59 fps frames.

This could be about a 50% performance gain with the relatively common native 29 fps videos when double rate deinterlacing is enabled. Another large performance gain could be checking 50 and 59 fps for film. It's quite common to see film in 720p59. Just giving madshi some ideas if he hasn't already thought about it. i can provide samples when the time comes.

[StinDaWg]

I've been messing with the settings trying to get the best setup for 720p->1080p NNEDI3 doubling, and I was noticing something was off but I couldn't pinpoint it until now. No matter what algorithm I use for downscaling, I get aliasing with text if anti-ringing is used. Anything else and text looks great, even Lanczos4 (no AR). Has anyone else noticed this or knows why this happens? I don't see any more ringing with it turned off, so I'm just going to leave it off for now.