[mythtv-users] mythcommflag logo detection does not work

[Brian J. Murrell]

On Mon, 2013-12-23 at 22:34 -0500, Michael T. Dean wrote:

> The best I can tell you is that commercial detection worked extremely 
> well in the US when it was developed (back in the days of SDTV) because 
> it was developed by a developer from the US.  It has progressively 
> gotten less reliable in the US because of changes to how broadcasters 
> send out the video--with different broadcasters doing things differently 
> resulting in varied success across different channels.

You know, everyone has this feeling that the commercial detection has
gotten worse over time, through different versions.  I'm in that boat.

I suspect that it's only related to time/versions in-so-much as how
things have changed since commflagging worked really well.  My gut tells
me that the most significant change over that time is move away from
encoding to MPEG2 from analog signals on capture cards such as the
Hauppage PVR-{{1,2}50,500) series cards towards what a lot of us do now
and that's direct capture of MPEG2 from the broadcaster.

When all that mythcommflag had to deal with was the MPEG2 that those
Hauppage cards produced, it was a very well known quantity.  Now
mythcommflag (well, ffmpeg, I suppose) has to deal with whatever funky
encoding might be coming from the provider directly and I suspect it's
not doing so well at that -- understandably.

The reason I have this feeling in my gut is because I have both types of
encodings here.  I have the QAM that my cableco sends me and I also have
a couple of those Hauppage encoder cards and I get recordings from both.
The commflagging on recordings from the Hauppage cards is still pretty
damn good, whereas the commflagging on the QAM recordings is not nearly
as good.

It's strange though, given that ffmpeg can decode these QAM captures
well enough to display them for our eyes but something about them makes
commflagging less accurate.  It would seem that if it could do one it
could do the other.  Which is why I just call this a "gut feeling".  I
have no science whatsoever beyond the anecdotal evidence gathered here.

I have had a long-time desire to break into the commflagging code to
debug this particular issue, but as Gary later posts -- life happens,
and seems to take up more and more of your (non-leisure) time the older
you get.

But I'd kill for accurate commercial detection since that's probably the
one deal-breaker that's preventing me from using XBMC as my FE over
MythTV[1].  XBMC only has two commercial detection handling modes: none
and automatic.  It doesn't have the middle-ground "skip commercial
buttons" handling that MythTV's FE does and using automatic with XBMC
and inaccurate detection is just frustrating and using none is a
non-starter, around these parts at least.

Cheers,
b.

[1] I still believe that if all of the MythFE development effort were
moved into XBMC, we'd have an overall better FE given that the overall
FE effort would be bigger where regular XBMC devs can concentrate on
their parts (regular videos, network streaming, etc.) and the MythFE
developers can focus their effort on purely playing Myth recordings and
things associated -- like adding commercial skip button handling --
instead of also having to handle video, network streaming, a browser,
weather plugin, etc.  But that's just my uninformed, outsider view of
things.  :-)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: This is a digitally signed message part
URL: <http://www.mythtv.org/pipermail/mythtv-users/attachments/20131224/4102e435/attachment.sig>