[mythtv-users] commercial flagging only some recordings

[Matt Doran]

Robert Tsai wrote:
> On Wed, May 23, 2007 at 08:52:22AM -0400, Michael T. Dean wrote:
>   
>> On 05/23/2007 04:36 AM, Leigh Porter wrote:
>>     
>>> However.. I am in the UK and I have noticed that commflag does not
>>> seem to work too well on UK channels, is this the case?
>>>       
>> Robert Tsai has been doing some work--in what free time he has--on
>> improving the flagger, especially for those outside the US.  And,
>> IIRC, he's in the UK, so once it's incorporated into Myth...
>>     
>
> The code is already accessible in MythTV as the "Experimental" option
> in the choice of commflaggers.
>
> I live in the US.
>
> I find that it works better than the "all" flagger, for me, for my OTA
> HD recordings.
>
> I can't say if it represents a regression from the "all" flagger for
> other kinds of recordings (cable TV, non-US, etc.). But if the "all"
> one doesn't work for you, then you may as well give "experimental" a
> shot.
>
> If you want to know why the "experimental" flagger isn't doing a good
> job, this page will describe how you can get a graphical depiction of
> what it thinks is going on:
>
> 	http://www.mythtv.org/wiki/index.php/User:Rtsai1111
>
> It is heavily biased towards logo detection (since that is a good
> indicator for OTA HD broadcasts, probably not as reliable for cable
> TV). If logo presence is not a good indicator in the UK, then this
> will probably not work so well for you.
>
>   
Hi Rob,

I'm in Australia, and I've found the default commflagger to be pretty 
hit and miss.  At a guess it probably gets less than 30-50% of 
commercials, but some channels are better than others.

All the stations in Aus make use of station logos/watermarks, so I'm 
going to give your experimental flagger a try.  I'm not sure what logos 
are like in other countries... but here in Aus they are quite 
transparent, and appear in different positions on each channel.  Most 
are in the bottom right, but at least one channel is in the upper right.

I'm curious to know how this differs from the default flagger, which 
also does logo detection.  Is it described anywhere?  Can you give us a 
bit of an overview?

Under what circumstances does the experimental flagger outperform the 
default one?

Thanks,
Matt