Dogway's Filters Packs

[pinterf]

Originally Posted by Dogway

But I will wait until you add selective frameprop copy if you still plan to add it.

Yes, probably next week second half.

[salvo00786]

Hi...
TDS, thanks for your reply.
This is my complete script

Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\vinverse\vinverse.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\SysInfo0129\SysInfo64.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\SysInfo0129\SysInfo32.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\vsTCanny\vsTCanny.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\RgTools12\RgTools.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\NNEDI309460\nnedi3.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\masktools22230\masktools2.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\mvtools2745\mvtools2.dll")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\SMDEGRAIN347d.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\Zs_RF_Shared1157.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\ExTools.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\QTGMC.avsi")
video=AssumeTFF(video).QTGMC(Preset="Slower")

I can tell you that all these strings are correct because i have all these plugin in the directories that you can see in my script.

1)I renamed vsTCanny-1.1.5 to vsTCanny, SysInfo_0129 to SysInfo0129.
2)All my scripts are in the scripts directory, and not in extra.
3)The new QTGMC in QTGMC.avsi, if I write QTGMC.avs, the program use the old QTGMC integrated.

[salvo00786]

Ok. Now it works. The resolutive step was the installation of the avisynth, thanks.

Another question. When I copied the dlls, for the 64bit dlls I used the dlls in the x64_CUDA_enable directory instead of the x64 directory, because I have a RTX3070 card. Is this correct?

[TDS]

Originally Posted by salvo00786

Hi...
TDS, thanks for your reply.
This is my complete script

Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\vinverse\vinverse.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\SysInfo0129\SysInfo64.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\SysInfo0129\SysInfo32.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\vsTCanny\vsTCanny.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\RgTools12\RgTools.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\NNEDI309460\nnedi3.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\masktools22230\masktools2.dll")
Loadplugin("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\mvtools2745\mvtools2.dll")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\SMDEGRAIN347d.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\Zs_RF_Shared1157.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\ExTools.avsi")
Import("A:\PORTABLE\RipBot264v1.26.0\Tools\AviSynth plugins\Scripts\QTGMC.avsi")
video=AssumeTFF(video).QTGMC(Preset="Slower")

I can tell you that all these strings are correct because i have all these plugin in the directories that you can see in my script.

1)I renamed vsTCanny-1.1.5 to vsTCanny, SysInfo_0129 to SysInfo0129.
2)All my scripts are in the scripts directory, and not in extra.
3)The new QTGMC in QTGMC.avsi, if I write QTGMC.avs, the program use the old QTGMC integrated.

OK, if that's how you want to have it setup, that's entirely up to you.

I use the "extra" folder, as I have added so many other scripts to RipBot, some of them have the same name, and they can't be in the same folder.

I don't use avsi, they should be avs, for RipBot, but again, if they work for you...good

As for QTGMC, you'll have to put it where the "old" one is, just rename the original, so that the process uses Dogway's build.

[TDS]

Originally Posted by salvo00786

Ok. Now it works. The resolutive step was the installation of the avisynth, thanks.

Another question. When I copied the dlls, for the 64bit dlls I used the dlls in the x64_CUDA_enable directory instead of the x64 directory, because I have a RTX3070 card. Is this correct?

So it obviously needed the newest version

I don't use the CUDA version, but if it works for you, then why not...you can always "test" the non CUDA ones.

I did ask that very same question, I'll see if I can find it, and let you know.

https://forum.doom9.org/showthread.p...85#post1955585

[Dogway]

Originally Posted by TDS

Error with ExTools v7.6, from v7.5, with DFTTest (for prefilter=4)

Error at Line 251 in SMDegrain v3.4 7d

Yes, noticed yesterday, a stupid typo in ExTools, updating later today with some other features.


EDIT: Updated ExTools, besides the fix I improved ex_corner() temporal coherence for that I needed a motion mask, since I didn't want to go to the MVTools route I made a new motion mask filter called ex_occlusion() works well as a MMask replacement and it's very fast.

[gispos]

@Dogway, please give your versions different names than the 'original'.

For example In your version of 'QTGMC' other parameters and functions are used. And it's also kind of misleading (what am I using right now).
And also the sub functions should be renamed from QTGMC.

I also can't create a function definition in AvsPmod for two times QTGMC, so I can only ban your version completely.
As default I use the normal QTGMC, it is faster for me.

I am so cheeky. Dogway's QTGMC renamed to ex_QTGMC, also sub functions. (2. upload)
A few places sub function rename forgotten and AvsPmod function definetion added.
https://www.mediafire.com/file/id2q3..._QTGMC.7z/file

[kedautinh12]

QTGMC changed name to STGMC by real.finder. I think you need change your QTGMC to STGMC. Why you need the developer of QTGMC from 2015 like Dogway need change his QTGMC to other name?? I think the person need change QTGMC to other name is you
https://github.com/realfinder/AVS-St...0up/STGMC.avsi

[Dogway]

@gispos what version is your 'original' QTGMC? My version is called QTGMC mod, you can name it like so since old version is outdated already (circa 2011) so I don't think there should be a name conflict. In 2012 I implemented 16-bit processing with Dither Tools and from 2015 real.finder took over, maybe you want to use STGMC?

The main reason Expr() is slower than MaskTools is explained in this post. The reason it's almost on par with real.finder QTGMC is because I optimized the filter to the extreme (merging expressions, wiser maths), same for LSFmod and other mods.

[TDS]

Originally Posted by Dogway

Yes, noticed yesterday, a stupid typo in ExTools, updating later today with some other features.


EDIT: Updated ExTools, besides the fix I improved ex_corner() temporal coherence for that I needed a motion mask, since I didn't want to go to the MVTools route I made a new motion mask filter called ex_occlusion() works well as a MMask replacement and it's very fast.

Nice, fixed, thanks.

Not sure what to do with all the other options within your scripts.

[Dogway]

Originally Posted by TDS

Nice, fixed, thanks.

Not sure what to do with all the other options within your scripts.

What options?

[TDS]

Originally Posted by Dogway

What options?

Maybe "options" was the wrong term, but if I look thru all the many, many lines of code, etc, there just seems (to me) that there are other filters & codes that maybe used as a stand alone.

eg:- ex_corner() and ex_occlusion()

or are these just part of the "big picture" of the main script ??

Forgive me, this is the "noobie" in me coming out.

[gispos]

Originally Posted by kedautinh12

QTGMC changed name to STGMC by real.finder. I think you need change your QTGMC to STGMC. Why you need the developer of QTGMC from 2015 like Dogway need change his QTGMC to other name?? I think the person need change QTGMC to other name is you
https://github.com/realfinder/AVS-St...0up/STGMC.avsi

Originally Posted by Dogway

@gispos what version is your 'original' QTGMC? My version is called QTGMC mod, you can name it like so since old version is outdated already (circa 2011) so I don't think there should be a name conflict. In 2012 I implemented 16-bit processing with Dither Tools and from 2015 real.finder took over, maybe you want to use STGMC?

The main reason Expr() is slower than MaskTools is explained in this post. The reason it's almost on par with real.finder QTGMC is because I optimized the filter to the extreme (merging expressions, wiser maths), same for LSFmod and other mods.

I'm probably behind the times. But it completely eludes me the logic.
QTGMC has been renamed to STGMC and the Dogway's full modeled version is now called QTGMC. What a mess.

I have hundreds of scripts with the original videos on CD and DVD, all old movies, many scripts contain QTGMC...


My version.

#-------------------------------------------------------------------#
#                                                                   #
#                     QTGMC 3.33 by Vit, 2012                       #
#     QTGMC 3.382 2021 mod by A.SONY, based on v3.33d by Dogway     #
#                                                                   #
#   A high quality deinterlacer using motion-compensated temporal   #
#  smoothing, with a range of features for quality and convenience  #
#          Originally based on TempGaussMC_beta2 by Didée           #
#                                                                   #
#-------------------------------------------------------------------#
#
# Full documentation is in the 'QTGMC' html file that may comes with this script, or even better see avisynth wiki
#
# --- LATEST CHANGES ---
#
# v3.382s
# - fix for n16 family
#
# v3.380s based on v3.34d (mod) by Dogway
# - Added Refinemotion Option for more accurate denoising.

This is from version v3.382s, Information is missing in STGMC 3.385, Has something changed?

# --- REQUIREMENTS ---
#
# Input colorspaces: YV12, YUY2, others as plugins supports
#
# Core plugins:
#	MVTools2 (2.7.26 or above)
#	MaskTools2 (v2.2.17 or above)
#	NNEDI3 (0.9.4.6 or above)
#	RgTools
#
# Core scripts:
#	Zs_RF_Shared.avsi
#
# Additional plugins and scripts:

Short test with STGMC went well, or are other things needed after all? How many children did QTGMC get now, is there also a QTGMCmod?

[Dogway]

Originally Posted by gispos

Short test with STGMC went well, or are other things needed after all? How many children did QTGMC get now, is there also a QTGMCmod?

Two only, three if I change to QTGMCmod.

# QTGMC 3.382 2021 mod by A.SONY, based on v3.33d by Dogway #

QTGMC mod is my branch for HBD (and now also for frameprops plus other), and indeed if you compare STGMC with QTGMC mod my version is way faster.
I don't claim my branch to be the official Vit's QTGMC replacement, so if you are very bothered by it and still want to use the 2011 version I can rename it to QTGMCM(). But as I say, for 8-bit QTGMC mod is not slightly slower than STGMC because I did something wrong in the code, if it's slower report upstream to improve Expr().

Also function names weren't a problem back then, you used the same name as a drop-in replacement, specially if the filter development ceased, that was in fact the intention so all your hundreds of old scripts wouldn't run into issues. As far as I'm concerned real.finder can still use QTGMC() or SMDegrain().

If you insist I should change the function name of most my filters, namely: GrainFactory3mod, SMDegrain, DestripeH and V, LSFmod, GMSD, MDSI, vsSSIM, BBSIM, SSIM_downsample, _IQA_downsample, RoundHalfToEven, Deblock_QED, MSUDeCor, SmoothDeblockn, SmoothD2c, ChromaReconstructor, ChromaReconstructor_faster, SPresso, SeeSaw, SeeSawMulti, FineSharp, NonlinUSM, SlopeBend, DetailSharpen, MedianSharp, SSSharpFaster, SSSharp, ReCon, blah, MedSharp, SharpenComplex2, pSharpen, CASm, Soothe and mmod (probably more)

Originally Posted by TDS

Maybe "options" was the wrong term, but if I look thru all the many, many lines of code, etc, there just seems (to me) that there are other filters & codes that maybe used as a stand alone.

eg:- ex_corner() and ex_occlusion()

or are these just part of the "big picture" of the main script ??

What can you use in replacement for ex_corner()? I wouldn't make something if I didn't see a value. AFAIK there's no corner detection filter in either AVS or VS (unless you use external libraries like OpenCV). As for motion masks, mt_motion/ex_motion is too primitive and MVTools' MMask() is too slow (5~15fps), ex_occlusion() runs at 470fps and worked well for my purpose, still it's first version so it can be improved.

If you don't write script filters ExTools might not be of much interest to you, it's basically a backend library to feed higher level filters. Readily usable filters are in EX/MIX mods or my other Packs.

[TDS]

Originally Posted by Dogway

What can you use in replacement for ex_corner()? I wouldn't make something if I didn't see a value. AFAIK there's no corner detection filter in either AVS or VS (unless you use external libraries like OpenCV). As for motion masks, mt_motion/ex_motion is too primitive and MVTools' MMask() is too slow (5~15fps), ex_occlusion() runs at 470fps and worked well for my purpose, still it's first version so it can be improved.

If you don't write script filters ExTools might not be of much interest to you, it's basically a backend library to feed higher level filters. Readily usable filters are in EX/MIX mods or my other Packs.

Hi Dogway, No, I don't write script filters, because I wouldn't know where to start, but I doubt that I need to do anything like that, for what I do.

I use ExTools because the other filters, etc, that you create, this is a major dependency, so I just go with what works for me.

Why I asked the question was simply because I see all these other "functions" within SMDegrain, ExTools (for example), and I just wondered if they could be used individually, if I knew how to....that's all.

Sorry for the frustration, just keep doing what you're doing, it's awesome

Cheers

[gispos]

Originally Posted by Dogway

If you insist I should change the function name of most my filters, namely:

I don't want to answare directly to that.
Suppose I want to use the 'original' filter and not your version, the reason should not matter now.

I load my filters via an avsi, so for example NonlinUSM is also loaded. Now I make my settings, ok, try a filter from Dogway.
What happens now? My NonlinUSM will be replaced by your version as soon as I load your SharpenerPack.
And your NonlinUSM version gives other results.

Possibly other values or parameters are necessary to achieve the same result (must not, but can be).
This is a no go for me.

Using the same names leads to confusion and is misleading.

[coolgit]

Originally Posted by gispos

Using the same names leads to confusion and is misleading.

I have to agree with Gispos here.

I have used some of your filters and it clashed with some filters already in my plugins folders and had to fanny around sorting it out. Eventually I had to remove them.

When newbies search for avisynth they will end up at this main official site
http://avisynth.nl/index.php/Main_Page
then
http://avisynth.nl/index.php/QTGMC
then
click on QTGMC.avsi and get 404 error. One can get to realfinder's page and see STGMC and either get confused or download it.

Correct me if i am wrong, the main difference between current QTGMC and yours is yours has high bit depth (HBD).
If so then why not use QTGMC_HBD like Stainlesss did with ScSelect_HBD. https://forum.doom9.org/showthread.php?t=182392

Realfinder can carry on using QTGMC. Avisynth site just need to add your filter in external filter page with documentation. That why it would be neat and tidy and won't scare the newbies away.

[real.finder]

my QTGMC (STGMC) also work with HBD

I don't mind if dogway keep the name if the function give same output as old one (which I think it do aside from SMDegrain which is not)

[coolgit]

OK that's good, thought it was 8 bit only.

The main avisynth site weblink needs to change QTGMC to STGMC and a new page for Dogway QTGMC. That's assuming the 'how to use documentation' is different.

Just checked your STGMC and can see it has been updated a bit.

[Dogway]

Originally Posted by gispos

I don't want to answare directly to that.
Suppose I want to use the 'original' filter and not your version, the reason should not matter now.

I load my filters via an avsi, so for example NonlinUSM is also loaded. Now I make my settings, ok, try a filter from Dogway.
What happens now? My NonlinUSM will be replaced by your version as soon as I load your SharpenerPack.
And your NonlinUSM version gives other results.

Possibly other values or parameters are necessary to achieve the same result (must not, but can be).
This is a no go for me.

Using the same names leads to confusion and is misleading.

If it gives other results REPORT. Must of my ports are meant to sanitize and unify code not to redefine its purpose (unless inconsistency). Why do you need 2 or 3 versions of NonlinUSM? A few handful versions of mod4(), two versions of Destripe or SPresso, 7 versions of Deblock_QED(), and so on... See my point? that leads to confusion and is misleading.

You need to understand what I'm doing. Why do you need:

"x x y - x y - abs 0.0001 + / *"

to change sign when you can now do

"x dup y - sgn *"

Same with the hundreds and inefficient ways to blur (I have even seen blur(1).blur(1).blur(1).blur(1).blur(1)...), like in your HighlightLimiter function.
NonlinUSM uses an unpredictable and non-linear way of blurring (what is 'rad' here anyway?):

bicubicresize(round(width()/rad/4)*4,round(height()/rad/4)*4).bicubicresize(width(),height(),1,0)

blurring more when small? Shouldn't you want to preserve the intention regardless of resolution? It should be the other way around (it was also reported earlier). For that I need to go back to the original thread and find out what resolution it was tested on and fit a function to scale that to any resolution.
In this case I ran into several issues because I developed NonlinUSM at the same time I worked on ex_GaussianBlur, but this was my take (it will never match the above SoftCubic100 upscale but the goal is to minimize error):

ex_GaussianBlur(rad*0.83825-0.0631,pad=true)

To make it work at any resolution simply scale rad based on it

ex_GaussianBlur((w/720.)*rad*0.83825-0.0631,pad=true)

Another point is filters with different arg names for the same purpose. 'show', 'showsettings', 'debug'... I do unify those so you don't have to look up the function arg list to find out. Same with 'chroma', bool 'UV', int 'UV', etc. Others like 'th', 'thr', 'thres', 'threshold'... or 'lo' and 'hi' in my filters which are a staple instead of 'low' and 'high', or 'th1' and 'th2', 'b' and 'w', 'dark' and 'bright', etc.

So the question goes back to, why do you want to use old, unmaintained and slow filters and what can we do to improve it?

I'm open to change the name of those that intentionally changes the output though (despite for fixing or improving purposes).

Originally Posted by real.finder

my QTGMC (STGMC) also work with HBD

I don't mind if dogway keep the name if the function give same output as old one (which I think it do aside from SMDegrain which is not)

Didn't mean that, my point was (replying gispos) that despite QTGMC mod being a tiny bit slower than STGMC at 8-bit it's way faster for HBD which was it's main purpose for the mod to start with. I changed a few output related things some months back and plan to do further with Zopti findings so I'll be doing more changes.

I will change the name of QTGMC and LSFmod to start with since I added critical output improvement. Probably with 'plus' as suffix since now it supports also frameprops aside HBD.