[mythtv-users] Commercial Flagging in 0.22 - no success.

[Corne Beerse]

Manuel McLure schreef:
> On Fri, Feb 26, 2010 at 11:10 PM, Manuel McLure <manuel at mclure.org> wrote:
>   
>> Well, turning off comflagging while recording doesn't seem to have
>> made a difference. I did notice that my SD recordings which are MPEG2
>> seem to commflag just fine, while with the HD recordings (h.264) I end
>> up with total garbage. I almost wonder if the detection is correct but
>> for some reason the written offsets are wrong...
>>
>> I've switched from "All Methods" to "Blank Frame + Scene" and I'll see
>> if any future recordings seem to flag better.
>>     
>
> OK, I did an experiment. I took a half hour recording, skipped through
> the flags, and wrote down the times for each skip. Then I went through
> the program skipping by 30-second intervals looking for the commercial
> breaks and wrote down the times for the start and end of each
> commercial break. I ended up with the following extremely interesting
> data:
>
> Real Time      Skipped Time     Ratio
>  7:18 (438)     5:37 (337)      1.3
> 11:19 (679)     8:53 (533)      1.27
> 16:51 (1011)   13:12 (792)      1.28
> 18:52 (1132)   14:47 (887)      1.28
> 24:18 (1458)   19:02 (1142)     1.28
> 26:20 (1580)   20:37 (1237)     1.28
>
> Real Time is the actual location of the start or end of the commercial
> break, skipped time is the place MythTV skipped to. The number in
> parenthesis is the time in seconds, and the ratio is the real time in
> seconds divided by the skipped time in seconds. Note that the saved
> skips are off by a factor of about 1.3 to where they really should be
> - it looks like it's correctly detecting the ads but somehow scaling
> the frame number before writing it to the database. Maybe an artifact
> of the h.264 encoding used?
>
> I checked the recordedmarkup database entry for the last skip and it's
> 37081 which at 30fps is just about 20:37. The last entry in
> recordedseek for the same recording is 53836 which is just about right
> for a 30 minute show at 30fps.
>
> Any mythcommflag experts want to chime in?
>   
Strange/funny. I've something similar with cutpoints and then 
transcoding and/or creating dvd-archives.

It is true that commercial flagging was way-of, specially further into 
the recording. I just updated them to make better cut-points. Then I 
started transcoding and/or creating dvd-archvies.

Then I found the archives and dvd-archives where way off: commercials 
where still in there but other parts where missing.

I expect it is not the commercial flagging, it is the representation of 
the cutpoints in the frontend!!!

btw: I'm located in The Neterlands, Pal-bg country with 50 Hz refresh or 
25 frames/second interlaced.

I might have an installation of ubuntu 8.4 somewhere around, with valid 
(checked, controlled and working) cut points. If time comes by, I might 
update that to the current ubuntu (that's relative easy) and see what 
happens with the cut points.

Mean while, how/where can we send in bugs on this? or is this message 
enough?


CBee