Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Capturing and Editing Video > Avisynth Usage

Reply
 
Thread Tools Search this Thread Display Modes
Old 12th April 2011, 09:20   #41  |  Link
Didée
Registered User
 
Join Date: Apr 2002
Location: Germany
Posts: 5,389
Quote:
Originally Posted by SubJunk View Post
I have added the new parameter "GPU" ...
Please let me know if it works for you
With GPU=false it doesn not work correctly for me.

On Intel Pentium-D (WinXP): the output shows a 2x2 array of the input.

On Intel i7-860 (Win7): every few moments, there are green "flashes" all over the screen. (Didn't look close - could be the green builds a "cross" pattern, related to the grid of the 2x2 array.)

Code:
src = colorbars().converttoyv12().reduceby2().showframenumber()
dbl = src.interframe(GPU=false,FlowPath="C:\ [...] \InterFrame-1.5\Dependencies\")
stackhorizontal( interleave(src,src).subtitle("source").addborders(4,4,4,4),
 \               dbl.subtitle("InterFrame(GPU=false)").addborders(4,4,4,4))
return(last)


Colorbars is just for demonstration. Running a native 1280x720 clip shows the same issue.
__________________
- We´re at the beginning of the end of mankind´s childhood -

My little flickr gallery. (Yes indeed, I do have hobbies other than digital video!)
Didée is offline   Reply With Quote
Old 12th April 2011, 09:23   #42  |  Link
Gavino
Avisynth language lover
 
Join Date: Dec 2007
Location: Spain
Posts: 3,431
Quote:
Originally Posted by -Vit- View Post
One thing that immediately stands out is the EEDI2 custom pelclip. That's from the MVTools docs; doesn't that example assume interlaced source? Because it discards half the lines, interpolates them back, then resizes.
EEDI2 doesn't discard half the lines. It resizes an image by 2x in the vertical direction by copying the existing pixels and interpolating new ones in between them. So it works for progressive sources - the MVTools doc example is assuming progressive as it doesn't call SeparateFields before EEDI2.
__________________
GScript and GRunT - complex Avisynth scripting made easier
Gavino is offline   Reply With Quote
Old 12th April 2011, 09:48   #43  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Quote:
Originally Posted by Didée View Post
With GPU=false it doesn not work correctly for me.

On Intel Pentium-D (WinXP): the output shows a 2x2 array of the input.

On Intel i7-860 (Win7): every few moments, there are green "flashes" all over the screen. (Didn't look close - could be the green builds a "cross" pattern, related to the grid of the 2x2 array.)

Code:
src = colorbars().converttoyv12().reduceby2().showframenumber()
dbl = src.interframe(GPU=false,FlowPath="C:\ [...] \InterFrame-1.5\Dependencies\")
stackhorizontal( interleave(src,src).subtitle("source").addborders(4,4,4,4),
 \               dbl.subtitle("InterFrame(GPU=false)").addborders(4,4,4,4))
return(last)


Colorbars is just for demonstration. Running a native 1280x720 clip shows the same issue.
Darn, that happens for me too.
It is related to that dll (libflowsse.dll) because it even happens with a simple script like this:

Code:
svp_flow_lib = "C:\path\libflowsse.dll"

Super = MSuper()
backward = MAnalyse(Super, isb=true)
forward  = MAnalyse(Super, isb=false)

MSmoothFps(Super, backward, forward, num=48000, den=1001, algo=23)
I will contact the SVP guys, it is probable that it was an incorrect assumption of mine that libflowsse.dll would work the same as libflowgpu.dll.
SubJunk is offline   Reply With Quote
Old 12th April 2011, 09:56   #44  |  Link
-Vit-
Registered User
 
Join Date: Jul 2010
Posts: 448
Quote:
Originally Posted by Gavino View Post
EEDI2 doesn't discard half the lines. It resizes an image by 2x in the vertical direction by copying the existing pixels and interpolating new ones in between them. So it works for progressive sources - the MVTools doc example is assuming progressive as it doesn't call SeparateFields before EEDI2.
Ah yes, I forgot; the meaning of "field" is completely different between EEDI2 and [EEDI3/NNEDI2/NNEDI3]. (...)
-Vit- is offline   Reply With Quote
Old 12th April 2011, 10:12   #45  |  Link
Gavino
Avisynth language lover
 
Join Date: Dec 2007
Location: Spain
Posts: 3,431
Quote:
Originally Posted by SubJunk View Post
it is probable that it was an incorrect assumption of mine that libflowsse.dll would work the same as libflowgpu.dll.
Perhaps it would have been a good idea to test this before releasing it...
__________________
GScript and GRunT - complex Avisynth scripting made easier
Gavino is offline   Reply With Quote
Old 12th April 2011, 11:59   #46  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Quote:
Originally Posted by Gavino View Post
Perhaps it would have been a good idea to test this before releasing it...
I had tested it, but not thoroughly enough. I had run a preview of the script and it was perfect, because the bug only affects 1 in every 2 frames (the preview happened to be on a good/original frame)
I would usually test more thoroughly but I had several users without GPUs posting that it no longer worked so I wanted to release the fix ASAP for them

Quote:
Originally Posted by -Vit- View Post
Ah yes, I forgot; the meaning of "field" is completely different between EEDI2 and [EEDI3/NNEDI2/NNEDI3]. (...)
That's convenient! I've also noticed that sometimes the documentation for those plugins just says "this has the same parameters as *function in other plugin*" so you have to open the other documentation It's a fun ride, but worth it

Last edited by SubJunk; 12th April 2011 at 12:04.
SubJunk is offline   Reply With Quote
Old 12th April 2011, 22:46   #47  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Version 1.5.1 is released with the GPU=false fix. Sorry to the people who encountered this bug.
SubJunk is offline   Reply With Quote
Old 13th April 2011, 06:25   #48  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Version 1.5.2 is released which is very minor, it just changes the default value of GPU from true to false because I discovered that the GPU mode reduces quality slightly.
It's a problem with SVP and the developers are aware of it so hopefully there is a fix coming soon
SubJunk is offline   Reply With Quote
Old 19th April 2011, 23:31   #49  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Version 1.6 is released which fixes the GPU mode bug and adds more presets for realtime users. Enjoy
SubJunk is offline   Reply With Quote
Old 20th April 2011, 22:00   #50  |  Link
jemm54
Registered User
 
jemm54's Avatar
 
Join Date: Feb 2011
Posts: 4
Hello, im having problems with this plugin, this is the error:

[Image]

im not sure what that means....

BTW, thanks for the plugin....

Edited: Make the picture smaller or upload it somewhere else and give a link. I've refused the pic.

Last edited by manono; 23rd April 2011 at 00:02.
jemm54 is offline   Reply With Quote
Old 20th April 2011, 22:34   #51  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Hi jemm You need a slash at the end of the FlowPath, so try:
InterFrame(FlowPath="C:\Archivos de programa\AviSynth 2.5\plugins\")

Last edited by SubJunk; 26th April 2011 at 21:14.
SubJunk is offline   Reply With Quote
Old 26th April 2011, 16:48   #52  |  Link
Gargalash
Registered User
 
Join Date: Nov 2008
Posts: 67
Hello Subjunk,
Thanks for this tool!

I have tried it and noticed something.

With default settings: --- EDIT: not default settings. I'm using preset "fast" and tuning="smooth" ---
- GPU= false, CPU usage is 48-50%, 42fps
- GPU= true, CPU usage is 46-48%, 46fps

The fps numbers are with vdub "video analysis pass" and my script has I have SetMTMode(2) before your filter.

If I don't use SetMTMode(2) I get about 13% CPU usage and 13fps for both true and false GPU. The AVI source I am using is on a SSD, I doubt it's creating the limitation.

Should this filter be using full cpu power? I have an i7 with multi-threading on.

Thanks for any input!
Gargalash is offline   Reply With Quote
Old 26th April 2011, 21:17   #53  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Hi Gargalash
Without SetMTMode the script will run in single-threaded mode, so 13% CPU probably just means it is using 100% of one CPU.
For the fastest results you should set the second number to 1.8x your thread count. So for example some quad-core CPUs have 4 cores but support 8 threads, so for that you would do 1.8*8 and round down to 14 so it would be:
SetMTMode(2,14)

Some people find they can just double the amount of threads and subtract one, but others find the 1.8x rule to provide better stability so I recommend that.

Hope that helps
SubJunk is offline   Reply With Quote
Old 27th April 2011, 06:04   #54  |  Link
PelPix
Registered User
 
Join Date: Feb 2009
Posts: 16
Hey SubJunk, I went back to InterFrame for motion blur because of the quality, but I'm having the problem I described earlier.
In the trails, the original frames are more visible than the interpolated frames. They have more weight in the accumulation/are held for longer/repeated more times. Is this intentional or a bug, and if it's intentional, is it for perceptual reasons?
PelPix is offline   Reply With Quote
Old 27th April 2011, 06:10   #55  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Hi PelPix Can you post an example so I can see what you mean?
SubJunk is offline   Reply With Quote
Old 27th April 2011, 06:11   #56  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Removed double-post
SubJunk is offline   Reply With Quote
Old 27th April 2011, 06:12   #57  |  Link
PelPix
Registered User
 
Join Date: Feb 2009
Posts: 16
Yup! one second!
I apparently just lost that source video and can no longer duplicate the problem . I'll let you know the second it comes up!
In other news, InterFrame+an accumulative temporal filter (Like TemporalSoften) makes a mean synthetic motion blur:

Yet another use for this really amazing script.
(Sorry about the quality. I had to screencap it from a render snapshot because it was currently rendering)

Last edited by PelPix; 27th April 2011 at 06:24.
PelPix is offline   Reply With Quote
Old 27th April 2011, 06:36   #58  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
It looks great, glad it's working for ya
SubJunk is offline   Reply With Quote
Old 27th April 2011, 18:08   #59  |  Link
Gargalash
Registered User
 
Join Date: Nov 2008
Posts: 67
Quote:
Originally Posted by SubJunk View Post
Hi Gargalash
Without SetMTMode the script will run in single-threaded mode, so 13% CPU probably just means it is using 100% of one CPU.
For the fastest results you should set the second number to 1.8x your thread count. So for example some quad-core CPUs have 4 cores but support 8 threads, so for that you would do 1.8*8 and round down to 14 so it would be:
SetMTMode(2,14)

Some people find they can just double the amount of threads and subtract one, but others find the 1.8x rule to provide better stability so I recommend that.

Hope that helps
Hi, thanks for your answer. I manage to get cpu usage to 93% and run the filter at 53-54 fps.
Gargalash is offline   Reply With Quote
Old 27th April 2011, 21:32   #60  |  Link
SubJunk
Registered User
 
Join Date: Jun 2010
Posts: 443
Quote:
Originally Posted by Gargalash View Post
Hi, thanks for your answer. I manage to get cpu usage to 93% and run the filter at 53-54 fps.
Great, I'm glad it works for you Remember that the presets control the speed and quality so you can make that FPS higher or lower with them.
Tunings also affect it; the smooth tuning will make it run slower.
SubJunk is offline   Reply With Quote
Reply

Tags
48fps, 60fps, framedoubling, interframe, smooth motion

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 12:45.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.