Commercial Advance Beta Release

[krutaw]

Since I've gone to the effort of running comskip on every video file I've got, I was thinking that it'd be great to come up with a way to be able to create a project file for Womble with the edits pre-determined based on the frames marked as commercials by comskip.

Alas, I've found that Womble doesn't keep track of the frames in the same ratio as comskip. I tried doing some math on the differences, but was unable to come up with a definitive difference between the two sets of numbers.

Anyone have any ideas on how Womble keeps track of it's frames?

[abasu2003]

Did anyone ever make a program to delete old comskip files?

[krutaw]

Originally Posted by abasu2003

Did anyone ever make a program to delete old comskip files?

Are you talking about any orphaned files? Yes, the vbscript that I wrote removes orphaned files, and after looking at Jere_Jones' automation script for Closed Captioning, it does as well.

As soon as I've tested my script a bit more, I'll make my vbscript available to everyone.

[ellweber]

Jere,

Thanks for fixing the file name problem, I've been out of town for a while. I now have a question about the logo detection. On some NBC shows ( and likely others as well) there is both a tranparent, relief style logo during the show and a full color logo that seems to be just about the same size. Can the logo detect differentiate? The full color version seems to show up in their ads for their own programming, which seems like a commercial to me! I've tried detect for "The Apprentice" both with and without the loge detect functions and the trial without seems much better though I believe that logo detection should enhance the performance.

Thanks for the hard work on this and the answers!

Lynn

[Jere_Jones]

Originally Posted by ellweber

Jere,

Thanks for fixing the file name problem, I've been out of town for a while. I now have a question about the logo detection. On some NBC shows ( and likely others as well) there is both a tranparent, relief style logo during the show and a full color logo that seems to be just about the same size. Can the logo detect differentiate? The full color version seems to show up in their ads for their own programming, which seems like a commercial to me! I've tried detect for "The Apprentice" both with and without the loge detect functions and the trial without seems much better though I believe that logo detection should enhance the performance.

Thanks for the hard work on this and the answers!

Lynn

Email me some screenshots showing the logos and I'll see. I know that normally (on SciFi and CBS, at least) the logos on the promos tend to be a little larger and a little higher, in which case comskip can tell the difference.

Jere

[ellweber]

Logo on the left is typical, on the right shown during ads for NBC programming. Taken from the same episode of "The Apprentice" and both referenced roughly to the lower right (oops!) corner of the screen.

Lynn

[Jere_Jones]

Comskip should have no trouble telling the difference. If it doesn't you can make comskip more strict via the ini file. Look at the logo_threshold property. It determines the percentage of edges that must match to be flagged as a logo.

Jere

[ellweber]

Thanks Jere,

I've added that parameter with the default value of 0.75 and am running it again. Report to follow.

Lynn

[heffe2001]

Has anybody else had crashing with .50 on SageTV 2.1? On my system it seems to crash about every time it's run (although I think I was having some problems with my recordings, what would a totall black screen recording do to comskip??)

[ellweber]

Looks much better now, guess I'm still finding my way through the many parameters!

Lynn

[Jere_Jones]

This is a bugfix release for those that have had comskip crash on certain mpegs. The only significant change is that comskip won't fail if there are no time stamps in the mpeg stream. I also beefed up the allocation routines to try to avoid using some unallocated space by accident.

Enjoy,
Jere

[why2kit]

Comskip just keeps getting better and better - keep up the good work ! On a side note does anyone know of software other than videoredo that can take the txt file and create an commercial free mpeg ?

[JasonJoel]

Just to let you know, COMSKIP is still crashing on me with v0.51.

It completed a few detections successfully last night, but when I came home from work today was crashed again.

Jason

Originally Posted by Jere_Jones

This is a bugfix release for those that have had comskip crash on certain mpegs. The only significant change is that comskip won't fail if there are no time stamps in the mpeg stream. I also beefed up the allocation routines to try to avoid using some unallocated space by accident.

Enjoy,
Jere

[Snuffy2]

Originally Posted by JasonJoel

Just to let you know, COMSKIP is still crashing on me with v0.51.

It completed a few detections successfully last night, but when I came home from work today was crashed again.

Jason

Exactly the same here.

End of crashed .log file looks like this:

Block 3 score:	Before - 2.00	After - 2.50	Combined count - 0
Block 4 score:	Before - 1.33	After - 1.67	Combined count - 0
Block 5 score:	Before - 1.00	After - 1.25	Combined count - 211
Combining Blocks 2 thru 6 result in strict standard commercial length of 119.45 with a tolerance of 0.767434.
Block 2 score:	Before - 2.50	After - 3.00	Combined count - 10
Block 3 score:	Before - 2.50	After - 3.00	Combined count - 0
Block 4 score:	Before - 1.67	After - 2.00	Combined count - 0
Block 5 score:	Before - 1.25	After - 1.50	Combined count - 211
Block 6 score:	Before - 1.00	Successfully deleted batchfile - comskip.bat

[Jere_Jones]

JasonJoel and Snuffy2, this will fix the crash that was demonstrated in the posted log. Keep up the bug reports! I really appreciate it.

Added Features

• added min_schange_modifier to ini properties
• added max_schange_modifer to ini properties
• csv out now outputs a histogram of each frame

Bugfixes

• fixed out of bounds errors during block combination

On a future development side, I'd like to explain the csv files and the reasoning behind them.
Comskip takes a long time to scan a recording. 99.999% of this time is used in decoding the mpeg stream. I had the idea that if I could save all the relavent information to a text file then I could try different settings quickly. That's where the csv file comes in. My preliminary testing shows that once the file is scanned, new settings take about half a second to be tested. During the writing of the code for this, it also pointed out how black frame detection is handled differently depending on if schene change detection is on or off. That will be fixed in the near future. Bear in mind that the csv functionality is NOT finished and may or may not work depending on the settings you use and the show you try it on. Regardless the results won't be accurate.

Enjoy,
Jere

[JasonJoel]

Yeah, I guess it would help if I posted the LOG file... DUH!!!!

Thanks, I'll give it a try and report back.

Jason

[ellweber]

I can't seem to send attachments to gmail accounts today so I guess this is the place. Sorry if this is duplication.

Hi Jere,

I have been having a little trouble lately with comskip. I thought I had it working quite well on Sunday but it has been down hill since. I was trying to adjust some of the parameters to improve the results but since that time I have been unable to get it to run to completion.

Three files are attached and I am hoping you will see the problem without much time required on your part. The log file and the comskip.ini used produce a blank .txt file. Seems to me like the process is stopping prior to the fuzzy logic part of the scene determination. I have upgraded the comskip.exe during this time and am now using 0.52. The other file is the .ini I was using before but it produces the same resulting blank .txt file. I tried the default .ini just to try to eliminate my tinkerings.

It is evident to me that you have put a lot of time and thought into this project and that you have some sophisticated insights into signal processing and TV practices. Don't ever think your work is not appreciated! Any clue as to what is going wrong here would sure help.

Thanks,

Lynn

[Jere_Jones]

Originally Posted by ellweber

It is evident to me that you have put a lot of time and thought into this project and that you have some sophisticated insights into signal processing and TV practices. Don't ever think your work is not appreciated!

I certainly appreciate the kind words. And at the very least, I appreciate my work. As far as sophisticated insights, I don't think so. I just know what I have learned from trying to improve comskip. Still, thanks.

Originally Posted by ellweber

Any clue as to what is going wrong here would sure help.

Unfortunately, I don't have a clue. The decoding just seems to stop. Very wierd. But, I'm attaching a special version that will give me very specific information about the exit that should help me. Run this one and send me the log.

Jere

[JasonJoel]

Well, 0.52 is still crashing on me. Here is a log file from the last file I tried to COMSKIP that caused it to crash.

Not sure if this is what you need to troubleshoot, if you need something else let me know.

EDIT: Ok more infromation.

1. That is DEFINITELY a file it crashes on every time
2. When it crashes, there are two COMSKIP.EXE files listed in task manager. One is ~5MB in size, the other is ~115MB in size.
3. There is a .TXT file generated, but it is 0 bytes, and empty.

Not sure if that helps you any.

Jason

[Jere_Jones]

JasonJoel,

Do you mean that it only crashes on one file?

Your log file ends the exact same way as ellweber's. Follow the same directions as above.

Jere