[mythtv-users] Commercial flagging on a sample basis

[Brad DerManouelian]

On Mar 25, 2008, at 6:10 AM, sylvain brejeon wrote:

> Hi all,
>
> Am I wondering if a commercial skipping method on a sample basis has  
> ever been attempted or discussed about here?

Yup. Check the archives. They've mostly been talked about in terms of  
sharing cutpoints in a peer-to-peer style network which ultimately was  
decided a very bad idea.

http://www.gossamer-threads.com/lists/mythtv/users/124644
http://www.gossamer-threads.com/lists/mythtv/users/85505

>
> A learning process in a way where some properties of commercials  
> would be recorded and then used for flagging.
> It wouldn't need a huge amount (to be updated overtime though) of  
> samples I think, as it is always the same bloody ads that come up  
> over and over again.
> Maybe even a database could be shared over the net.
> Please let me know if it already exists or what you think about it.
>
> Also I was thinking about the (US) series, aren't they formatted so  
> the commercial breaks last 3 minutes every 5 mins of actual programs  
> for instance?
> I'm not sure that it would be a viable method but it would permit to  
> release the CPU at regular intervals...
> well, was just a thought and maybe it's already the case with  
> current flagging methods.

That's a more interesting idea to me. If you consistently watch a show  
every day that has commercial breaks within the same couple of seconds  
in the same spots every night (I'm thinking Daily Show, Colbert Report  
for myself) then you could just re-use the commercial skip data from  
another program and apply it to the one that's just recorded.

I think the reason nothing like this has ever been implemented is  
because commercial detection is VERY fast. Faster than real-time  
video. As I'm watching a program, I already have commercial skip marks  
so it doesn't provide me very much functionality to not have to run  
the commskip process. Because of this, I would guess it makes more  
sense to put effort into making the commercial skip more accurate  
rather than effort into re-using less accurate existing commercial  
skipping. :)

-Brad