Detection near perfect, but goes really bad
Detection near perfect, but goes really bad
I've got comskip nearly perfect most of the time, but lately I've been seeing several occasions where it looks good while processing and everything gets screwed up just before it finishes processing. What is it enforcing at the very end?
Re: Detection near perfect, but goes really bad
I have a similar issue, when using -d 2
It doesn't want to build any commercial blocks (big gray blocks on the top line).
-d 10 doesn't help to make it understand that I only want to use logo as a detection method, and still have it create good blocks.
-d 11 (which includes uniform frames), does however allow me to see gray blocks representing the commericalbreak.
Is comskip uniformity oriented?
Because it seems that it will only process the final results, if it knows where the black frames are...
Problem is that one of the commercials is mostly black with some white text in the middle..
and so it stops the commercialbreak before that commercial, while infact it should continue for another 20s or so.
It doesn't want to build any commercial blocks (big gray blocks on the top line).
-d 10 doesn't help to make it understand that I only want to use logo as a detection method, and still have it create good blocks.
-d 11 (which includes uniform frames), does however allow me to see gray blocks representing the commericalbreak.
Is comskip uniformity oriented?
Because it seems that it will only process the final results, if it knows where the black frames are...
Problem is that one of the commercials is mostly black with some white text in the middle..
and so it stops the commercialbreak before that commercial, while infact it should continue for another 20s or so.
Re: Detection near perfect, but goes really bad
During live processing a simplified algorithm is used that does not use sound.
If the sound decoding fails you can have bad results.
Also the global heuristics only work when all processing is done.
Send me a .csv file, .log file and a .ref file as described in the "how to ask for help" topic in the support forum and I will have a look.
If the sound decoding fails you can have bad results.
Also the global heuristics only work when all processing is done.
Send me a .csv file, .log file and a .ref file as described in the "how to ask for help" topic in the support forum and I will have a look.