Thus said AJ ONeal on Thu, 15 Dec 2011 15:27:25 MST:

> I  thought  about using  imagemagick  to  compare differences  between
> frames and select the frames with the highest differences between them
> - but writing the scripts and code  to do that would have taken longer
> than the cumulative hour of watching the videos.

What  if you  write out  the  video to  .5  second frames,  then find  a
bayesian classifier  that can classify  images. Train the  classifier to
recognize images that have the features you're looking for and then when
you have fed  it enough samples, run  it over the whole lot  and have it
tell you which ones are potentially interesting. If it gets a lot wrong,
retrain, and try again.

Of course, by the time you get  all this working, you might already have
been able to watch the 10 hours. Silent movies used to be popular...

Andy


/*
PLUG: http://plug.org, #utah on irc.freenode.net
Unsubscribe: http://plug.org/mailman/options/plug
Don't fear the penguin.
*/

Reply via email to