Search found 116 matches

by kerryklarson
Sun Mar 04, 2007 1:08 pm
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Outstanding!! I ran through Heroes and Lost, and Comskip detected perfectly. However, American Idol and Best Week Ever make me think that the cut points are still 3 or more frames late. When shows have long black intervals, you can't detect small errors. However, there are some areas where the error...
by kerryklarson
Sun Mar 04, 2007 7:14 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

The "fuzzy logic" heuristics seems to be causing some problems. I think I see now why you mentioned this. In the log file is a section entitled: "Fuzzy scoring of the blocks". However, Fuzzy Logic was not one of the methods specified: Detection Methods to be used: 1) Black Frame...
by kerryklarson
Sun Mar 04, 2007 12:20 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Sometimes I check, but if there is no black frame sequence or if there is a series of dark frames, I don't worry too much about it since the problem is understandable. It's those cuts that are made right in the middle of normal program material that are frustrating. Fortunately, a lot of that is get...
by kerryklarson
Sat Mar 03, 2007 10:56 pm
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

I may have found the reason for the pre-logo/after-logo frame numbering difference. That's good news. A possible workaround is for you to set the PID in the commandline to 31 as all your TP recordings seem to use PID 31 for the video. I've been doing that for many months. I always specify the PID o...
by kerryklarson
Sat Mar 03, 2007 2:09 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

There is yet one other nasty problem, after comskip has find the logo for the first time it rewinds and start the processing from the start. Then, for some yet unknwon reason, the frame numbers can get a off by around 10 frames. Don't know yet how this happens but it should explain what you are see...
by kerryklarson
Fri Mar 02, 2007 11:23 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Erik, my sincerest apologies (again). I actually didn't know that you used the *.ref file in an automated way. I'm always looking at the videos in VRD, so the *.txt or *.ref files don't do me much good, only the *.vprj. I will freely admit to not understanding the magic you use to figure out these l...
by kerryklarson
Fri Mar 02, 2007 10:17 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Two steps forward, one back

Just when I think everything is tweaked out, I find new problems. See the log files at ftp://athome.kaashoek.com/Kerry/AI/. Frame 45211 Last black frame 45207 69286 Last black frame 69275 106540 Last black frame 106527 146054 Last black frame 146036 (however, only a single black frame) 180491 Last b...
by kerryklarson
Fri Mar 02, 2007 6:39 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

The first part of the show after the first commercial does not have logo. Thanks. I thought that might be the problem. However, I also thought that the distinct black frame transition and aspect ratio change, in addition to an excessively long commercial block with that part included, would have sc...
by kerryklarson
Fri Mar 02, 2007 2:10 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Erik, Could you take a look at the log files in ftp://athome.kaashoek.com/Kerry/Lost/. Comskip mistakenly included frames 34385-51215 as part of the previous commercial block and cut several minutes out of the show. I don't understand why since there was clearly a black frame sequence and aspect rat...
by kerryklarson
Fri Mar 02, 2007 12:33 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

YOur conclusion that repair=0 is consistent with VRD means indeed that VRD reads the file and presents the frame numbers in the UI WITHOUT using the PTS. COuld you post that conslusion at the VRD website? For anyone following this thread, the default behavior in VRD is to check and correct some err...
by kerryklarson
Thu Mar 01, 2007 11:51 pm
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

This was a clear example where scene_change (method=4) and uniform made things worse. I have not yet found good many good reasons to use method 4 . Try not to use this method. Indeed, deleting method 4 prevented that show segment from being cut. Looks like I'm down to method 35. Your zip file did n...
by kerryklarson
Thu Mar 01, 2007 1:36 pm
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Hmmm. I think I screwed up and SAVED the segment which means that VRD could have repaired the timeline. I've TRIMMED (no correction) the same segment and am uploading it now. It should finish in about 50 mins. In the meantime, I've uploaded log files from another show to the ICOE folder with a short...
by kerryklarson
Thu Mar 01, 2007 11:11 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Erik,

Please check the Wed Feb 28, 2007 7:50 am message in this thread. I posted two in a row, so you might have missed it. The files are there. Thanks.

Kerry
by kerryklarson
Thu Mar 01, 2007 10:49 am
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Closed Captions only works for recordings made with the Hauppauge PVR250 or PVR350 I did not find any CC in the recordings you did send me in the past. Something else good to know. I'm surprised because I thought that most ATSC prime-time broadcasts were closed captioned. Looks like I'm down to det...
by kerryklarson
Wed Feb 28, 2007 10:03 pm
Forum: Support
Topic: Inaccuracy
Replies: 83
Views: 53943

Sound is always used for detection of black frames and others, however "silence" is a detection method in its own using only the sound to find cutpoints. On films with a lot of silence this method certainly does NOT work. Good to know. Can you confirm that method 16, Closed Captions, work...