[mythtv] [mythtv-commits] Ticket #8522: Include only half the blank frames between commercial breaks

[Michael T. Dean]

  <moving conversation to the list>


On 09/11/2010 07:47 PM, Jim Stichnoth wrote:
> On Sat, Sep 11, 2010 at 11:38 AM, Michael T. Dean wrote:
>>   On 06/02/2010 02:07 AM, MythTV wrote:
>>> #8522: Include only half the blank frames between commercial breaks
>>>
>>> ------------------------------------------------+---------------------------
>>>   Reporter:  Jim Stichnoth<stichnot@…>             |       Owner:  cpinkham
>>>       Type:  enhancement                         |      Status:  new
>>>   Priority:  minor                               |   Milestone:  unknown
>>> Component:  MythTV - Mythcommflag               |     Version:
>>>   Unspecified
>>>   Severity:  medium                              |     Mlocked:  0
>>>
>>> ------------------------------------------------+---------------------------
>>>   When the commercial skiplist is created, I prefer to include a blank
>>>   interval between program sections (CommSkipAllBlanks=0).  Sometimes,
>>>   though, a small piece of the commercial break gets included.  A pretty
>>>   good compromise is to include half of the blank interval.  This patch
>>> adds
>>>   that as a CommSkipAllBlanks option.
>>>
>>>   Note: The patch addresses all places where the skiplist is adjusted
>>>   according to the blank sections, but I have only tested the portion in
>>>   BuildAllMethodsCommList().
>> Jim, this patch sounds interesting.  However, it also brings up another
>> possibility that could support our "get rid of some of the settings and use
>> reasonable defaults that work for everyone" initiative.
...
>> Would this approach work for you?  If so, would you be interested in working
>> on rewriting it to remove the CommSkipAllBlanks and just "do the right
>> thing"?  If not, I'll get to it eventually, but I just thought that since
>> you know what to look for in before/after behavior, you might do a better
>> job at testing and verifying the approach.
>>
>> It's probably not something that could go in for 0.24, though, so no hurry
>> on it.
> I think what we really need is more data.  I will modify my code base
> to log the boundaries of the blank interval to the recordedmarkup
> table.  Then after I fine-tune the cutlist, I can generate a report on
> the overall effectiveness of the commflagging, as well as details on
> the size of the blank interval and whether the fine-tuned cut mark
> falls nicely within the blank interval.  The latter will of course be
> biased by my choice of including half the blank interval (#8522).
>
> Honestly, though, I think the best one-size-fits-all method will be
> the #8522 option, i.e. setting the cut mark to the midpoint of the
> blank interval.  There are no magic constants to tune, (possibly
> changing) frame rates to consider, etc.  I'll have a better sense for
> this in two weeks, after the Fall season has been going in full for a
> week.  Of course, if this argument happens to convince you and others,
> I could save myself the analysis effort... :)


Yeah, I have no idea what the stats are like--how many frames should be 
blank and how many get seen as blank (especially in the cases of 
misdetection).  The idea of using half the frames all the time works for 
me, as long as we don't just end up using "half a runaway number."  
Maybe half with a upper bound for a sanity check?

It's probably easiest for you and Chris to hash out a plan on the list, 
so I'm moving this here.  Also, Chris may be able to provide info that 
allows you to skip the stats collection--or you guys may just decide on 
an approach that means it's not necessary.

Thanks,
Mike