DGAVCDec 1.0.9

[Jawed]

Would demuxing software make a difference? Should I upload the .h264 file that I extract from the MKV? Should I try the .h264 file that you extracted?

I have tried previewing the clip (F5) within DGAVCIndex at full resolution (Display HD Full Sized) and the artefacts also appear there.

Jawed

[Guest]

Originally Posted by Jawed

Would demuxing software make a difference? Should I upload the .h264 file that I extract from the MKV? Should I try the .h264 file that you extracted?

You should upload the source file. Don't make a new MKV out of anything. What did you download and do you have legitimate rights to have it?

I have tried previewing the clip (F5) within DGAVCIndex at full resolution (Display HD Full Sized) and the artefacts also appear there.

See, how can I do that with your sample??? DGAVCIndex does not open MKV.

Finally, post a screenshot showing the artifact.

[Jawed]

Originally Posted by neuron2

You should upload the source file.

It's 1.3GB.

Don't make a new MKV out of anything.

I originally made the MKV (from the .h264 extracted from the .flv) in order to extract 10+ seconds out of the source video. The artefacts are the same regardless of whether I muxed or not - and indeed the directshowsource version shows no artefacts, indicating that muxing is irrelevant.

What did you download and do you have legitimate rights to have it?

The filename tells you and I already said it came from iPlayer, which provides a 30 days licence to people like me who are UK residents.

See, how can I do that with your sample??? DGAVCIndex does not open MKV.

I was suggesting that I send you the .h264 that I extracted from the MKV. Or that you send me the .h264 you extracted. But I don't think that's going to lead to a solution anyway...

I made the MKV so that anyone else could compare directshowsource to avcsource with the minimum fuss as well as it being a convenient way to extract a few seconds showing the problem.

Finally, post a screenshot showing the artifact.

See two images here:

http://www.mediafire.com/?3dimyjznn2y

frames 240 and 241.

This is the script used to create the pix:

a=avcsource("Dan_Cruickshanks_Adventures_in_Architecture_-_Pleasure_b00brpqp_default-002_Track1.dga")
b=directshowsource("Dan_Cruickshanks_Adventures_in_Architecture_-_Pleasure_b00brpqp_default-002.mkv")
a=a.crop(400,300,-400,-200,align=true).subtitle("dga")
b=b.crop(400,300,-400,-200,align=true).subtitle("mkv") 
stackvertical(a,b)

I'm wondering if this has something to do with the GOPs. I'm not sure if this is the correct interpretation, but in DGIndex when I press the ">" button in the bottom right corner the frame-slider moves in steps and the second step appears to be the start of GOP 3 in this clip. It looks like it might be frame 241, which is the first frame that's artefact-free.

Thanks,
Jawed

[Guest]

I can't duplicate it. And those don't look like decoding errors anyway.

[poisondeathray]

Those errors are replicated in my testing with DGAVCDec 1.0.9, in frames beginning around 120 or so and ending at 240 (basically the entire zoom in on the black & white picture sequence), and the errors disappear on frame 241

Not present with ffms2() or directshowsource()/dss2() +ffdshow

[Jawed]

For what it's worth I'm running an Athlon X2 3500+ and detected SIMD is MMX, SSE MMX, SSE2, SSE FPU, 3D Now!

In general with this entire video I'm noticing that there are "snap changes" in the appearance of the video. I might upload a sample encode of a short sequence that shows this, later (after midnight my time). The comparison encode using DSS has a very slight "snap" change at the same time - much harder to see. I suppose this snap could be the start of a new GOP in the source.

Jawed

[Guest]

OK, I see it now. Seems to be caused by the old version of libavcodec that I am using. I cannot upgrade due to newer regressions and so cannot do anything about this.

It works fine with DGAVCDecNV.

[Jawed]

Thanks. I'll try FFMS2, never heard of it before (thanks poisondeathray for your report) - just to have something that avoids having to frobnicate FFDShow settings between encoding and viewing conditions.

Jawed

[poisondeathray]

Originally Posted by Jawed

Thanks. I'll try FFMS2, never heard of it before...

Sorry , to be more precise it's ffmpegsource2() or ffvideosource() , "ffms2" is the short name

http://forum.doom9.org/showthread.php?t=127037
http://code.google.com/p/ffmpegsource

[Jawed]

Thanks, I'd already completed a test encode, which I find preferable (less obnoxious blocking crud and barely noticeable "snap"), and will re-encode Dan Cruickshank's programme...

Jawed

[medp7060]

I just found that there is a huge difference for avcsource and directshowsource. Sample clip is here:

http://rapidshare.com/files/288791896/Flag.rar

In this case, avcsource screwed the clip while directshowsource handled it well. I used DGAVCDec Version 1.0.9.

LoadCplugin("C:\Program Files\AviSynth 2.5\plugins\yadif.dll") 
directshowsource("Flag.ts").AssumeTFF
#avcsource("Flag.dga").AssumeTFF
#avcsource("Flag.dga", deblock=false).AssumeTFF
Yadif (order=1)
crop(4,4,706,572).BicubicResize(704,528)

Edited: so either DGAVCDec or avcsource has a bug.

[halsboss]

dss2 ? directshowsource was thought to be less reliable than dss2 in some cases depending on source.

[Guest]

so either DGAVCDec or avcsource has a bug

PAFF is not supported by the early version of libavcodec I use in DGAVCDec, as mentioned in the readme.

The stream is properly handled by DGAVCDecNV.

[medp7060]

I see. Unfortunately, I cannot use it because

DGAVCDecNV 1.0.13: GPU decoding on Nvidia

--------------------------------------------------------------------------------

ATTENTION: If you do not use an Nvidia graphics card 8xxx or higher, this thread is not for you!

[b66pak]

hi,

beside indexing .h264 streams i use this tool for trimming...how can i see the number of the current frame when i move the slider?
_

[Guest]

You can't.

[Guest]

Originally Posted by neuron2

Thank you for notifying us.

The killer for me is Issue 475.

https://roundup.mplayerhq.hu/roundup/ffmpeg/issue475

I have several files that do this. As upgrading is a big effort, I prefer to wait until this major issue in PAFF support is fixed. Note that after more than 6 months it is still unassigned.

This is now shown as fixed so I will do some testing and if things are OK, I'll try to upgrade to the latest libavcodec, and thereby provide working PAFF support.

[Guest]

My initial test shows that the blocking issue is fixed!

So I'll make an updated version of DGAVCDec with the latest libavcodec.

[elguaxo]

[hydra3333]

Thanks !