Bug #1062
closedError creating Logo
Description
I get serious errors, when I try to extract the SAT.1-HD-logo. Here is the output:
root@ReelBox:/media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec# l
insgesamt 9,2G
drwxrwxr-x 2 root ftpusers 4,0K 2012-09-08 23:07 .
drwxrwxr-x 3 root ftpusers 8 2012-09-07 23:01 ..rw-rw-r- 1 root ftpusers 782M 2012-09-07 23:10 00001.tsrw-rw-r- 1 root ftpusers 2,0G 2012-09-07 23:35 00002.tsrw-rw-r- 1 root ftpusers 2,0G 2012-09-07 23:58 00003.tsrw-rw-r- 1 root ftpusers 2,0G 2012-09-08 00:19 00004.tsrw-rw-r- 1 root ftpusers 2,0G 2012-09-08 00:41 00005.tsrw-rw-r- 1 root ftpusers 574M 2012-09-08 00:51 00006.tsrw-rw-r- 1 root ftpusers 1,3M 2012-09-08 00:51 indexrw-rw-r- 1 root root 609 2012-09-08 23:06 inforw-rw-r- 1 root ftpusers 609 2012-09-07 23:11 info.txtrw-rw-r- 1 root root 16K 2012-09-08 00:51 preview_vdr.png
root@ReelBox:/media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec# markad --extractlogo=1 -v - /media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec
markad: [5081] starting v0.1.3 (32bit)
markad: [5081] on /media/hd/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec
markad: [5081] getting broadcast start from directory
markad: [5081] pre-timer 2m
markad: [5081] broadcast length 124m
markad: [5081] found H264-video (0x00ff)
markad: [5081] found AC3 (0x0103)
markad: [5081] using libavcodec.so.52.20.1 with 1 threads
markad: [5081] using codec H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10
markad: [5081] channel SAT_1_HD
markad: [5081] processing file 00001
[h264 0x8f583b0]number of reference frames exceeds max (probably corrupt input), discarding one 0x8f583b0]mmco: unref short failure
Last message repeated 393 times
[h264
markad: [5081] HDTV 1920x1080i25
[h264 0x8f583b0]number of reference frames exceeds max (probably corrupt input), discarding one 0x8f583b0]mmco: unref short failure
[h264
[h264 @ 0x8f583b0]number of reference frames exceeds max (probably corrupt input), discarding one
markad: [5081] finished logo extraction
root@ReelBox:/media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec#
The info file is:
root@ReelBox:/media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec# cat info
C S19.2E-1-1017-61300 SAT.1 HD
E 34606 1347050220 7440 4E D
T Das Schweigen der Lämmer
D Ein Serienkiller zieht seine blutige Spur durch den mittleren Westen der USA. Das FBI tritt auf der Stelle. Nur einer könnte der Truppe um Jack Crawford helfen: Dr. Hannibal Lecter. Der psychopathische Massenmörder und Psychologe sitzt in einem Hochsicherheitstrakt. Er soll ein Profil des Serienmörders erstellen und damit die Fahndung voranbringen. Crawford setzt die FBI-Schülerin Clarice Starling auf Lecter an, doch er weiß, dass Lecter nicht ohne Zugeständnisse helfen wird ...
X 2 05 deu
F 25
P 50
L 99
H
root@ReelBox:/media/reel/recordings/Das_Schweigen_der_Lämmer/2012-09-07.22.35.5-0.rec#
Also the pretimer of 2 minutes is not correct.
Files
Updated by Joe_D over 11 years ago
- Assignee set to Joe_D
Where are the 'serious errors'?
The H264-messages are from libavcodec, they will disapear when you omit the -v switch
Pretimer detection is ok cause your volume is not mounted with noatime option - time from directory is 22:35 (2012-09-07.22.35.5-0.rec) and time from info.txt (1347050220) is 22:37
Are you missing the extracted files in /tmp?
Updated by satellit over 11 years ago
Well, the errors will disappear without the -v option but libavcodec will have still the same problem, isn't it?
The logos are almost completely white and useless. Therefore I think that it is a serious error. Is this a memory issue?
The pretimer detection is not correct because it is a manual recording. You see this in the info file where the search timer information in the xml format is missing (see the tagesschau info file). I started the recording at 22:37 but the film started 2 minutes earlier. The info shows the time when I started the recording. The film started at 22:35 which is in the directory name and therefore I lost 2 minutes. The correct pretimer should have been - 2 minutes (minus two minutes) instead of + 2 minutes.
root@ReelBox:/media/reel/recordings/Tagesschau/2012-09-10.19.58.1-0.rec# cat info
C S19.2E-1-1019-10301 Das Erste HD
E 70315345 1347300000 900 0 FF
T Tagesschau
D Genre: Nachrichten|Die Nachrichten der ARD|Land: D|Stereo: Ja|Breitbild: Ja|Genre: Nachrichten
X 2 03 deu Stereo
X 1 03 deu 16:9
X 2 03 mis
X 2 05 deu
X 3 03 deu
F 50
P 50
L 99
@ <epgsearch><channel>1 - Das Erste HD</channel><update>0</update><eventid>70315345</eventid><bstart>120</bstart><bstop>600</bstop></epgsearch><pin-plugin><protected>no</protected></pin-plugin>
H
root@ReelBox:/media/reel/recordings/Tagesschau/2012-09-10.19.58.1-0.rec#
It seems to me that there is a discrepancy between search timer and manual recordings.
Thanks for your support!
Updated by Joe_D over 11 years ago
- Status changed from New to Assigned
- % Done changed from 0 to 50
satellit wrote:
Well, the errors will disappear without the -v option but libavcodec will have still the same problem, isn't it?
I don't know if these messages from libavcodec are errors or just warnings...
The logos are almost completely white and useless. Therefore I think that it is a serious error. Is this a memory issue?
No, try increase the size, e.g. --extractlogo=1,350,200 - but descrease the size afterwards to the smallest possible value...
The pretimer detection is not correct because it is a manual recording.
It seems to me that there is a discrepancy between search timer and manual recordings.
I must investigate this further. Thanks for the hint!
Updated by Joe_D over 11 years ago
- Status changed from Assigned to Feedback
- Priority changed from Normal to Low
- % Done changed from 50 to 80
The pretimer detection is not correct because it is a manual recording.
It seems to me that there is a discrepancy between search timer and manual recordings.I must investigate this further. Thanks for the hint!
I found no errors:
First an recording with pause/live-recording (markad recognizes the @-sign and disables pre-/post timer):
Tue Sep 11 12:59:32 [2315] starting v0.1.4pre (32bit)
Tue Sep 11 12:59:32 [2315] on /var/lib/video.00/@Malcolm_mittendrin_Verliebt_in_ein_Auto/2012-09-11.12.59.5-0.rec
Tue Sep 11 12:59:42 [2315] live-recording, disabling pre-/post timer
Tue Sep 11 12:59:42 [2315] broadcast aspectratio 4:3 (from info)
Tue Sep 11 12:59:42 [2315] broadcast with DolbyDigital2.0 (from info)
Tue Sep 11 12:59:42 [2315] broadcast length 28m
Tue Sep 11 12:59:42 [2315] found H262-video (0x01ff)
Tue Sep 11 12:59:42 [2315] found AC3 (0x0203)
Tue Sep 11 12:59:42 [2315] using libavcodec.so.52.20.1 with 2 threads
Tue Sep 11 12:59:42 [2315] using codec MPEG-2 video
Tue Sep 11 12:59:42 [2315] channel ProSieben
Tue Sep 11 12:59:42 [2315] SDTV 720x576i25
Tue Sep 11 12:59:42 [2315] no logo for ProSieben-A4_3
Tue Sep 11 12:59:42 [2315] aspectratio start with 4:3 (0)*
Tue Sep 11 13:07:47 [2315] aspectratio change from 4:3 to 16:9 (11772)
Tue Sep 11 13:08:32 [2315] aborted by user
C S19.2E-1-1107-17501 ProSieben
E 2501 1347361020 1680 4E E
T Malcolm mittendrin
S Verliebt in ein Auto
D Lois kommt völlig entsetzt [...]
G 14
X 1 01 deu 4:3
X 2 03 deu deutsch
X 2 05 deu Dolby Digital 2.0
F 25
P 50
L 99
Second example, broadcast made with VDR-EPG Recording function:
The dir-time 12:46 is misleading. It is NOT the real start of the recording! Start of the recording was 13:09, start of the event was 12:56 - so markad finds the correct truncation from the directory atime (only available on volumes with noatime mount option).
Tue Sep 11 13:09:20 [2407] starting v0.1.4pre (32bit)
Tue Sep 11 13:09:20 [2407] on /var/lib/video.00/Privatdetektive_im_Einsatz/2012-09-11.12.46.6-0.rec
Tue Sep 11 13:09:30 [2407] getting broadcast start from directory atime
Tue Sep 11 13:09:30 [2407] broadcast start truncated by 13m, length will be corrected
Tue Sep 11 13:09:30 [2407] broadcast length 42m
Tue Sep 11 13:09:30 [2407] found H262-video (0x00a6)
Tue Sep 11 13:09:30 [2407] using libavcodec.so.52.20.1 with 2 threads
Tue Sep 11 13:09:30 [2407] using codec MPEG-2 video
Tue Sep 11 13:09:30 [2407] channel RTL2
Tue Sep 11 13:09:45 [2407] SDTV 720x576i25
Tue Sep 11 13:09:45 [2407] detected logo start (27)*
Tue Sep 11 13:11:34 [2407] aspectratio of 16:9 detected[...]
C S19.2E-1-1089-12020 RTL2
E 90 1347360960 3360 4E 19
T Privatdetektive im Einsatz
D Detektei-Chef [...]
G 10 80 15
X 2 03 deu
F 25
P 50
L 99
X 1 03 deu 16:9
Updated by satellit over 11 years ago
Yes, the dir time 12:46 is misleading. I started this record via epg, the xml tags are missing and the epg offers as start time the start time of the film minus the default pretimer of 2 minutes: 19:10 - 2 min = 19:08
This is the dir time. The real start is about 19:45. The marks are correct.
root@ReelBox:/media/reel/recordings/Star_Trek_-_Deep_Space_Nine/2012-09-11.19.08.12-0.rec# cat marks
0:00:00.01 aspectratio start with 4:3 (0)*
0:38:47.07 stop of recording (58181)
root@ReelBox:/media/reel/recordings/Star_Trek_-_Deep_Space_Nine/2012-09-11.19.08.12-0.rec# cat info
C S19.2E-1-1109-5401 TELE 5 HD
E 70356392 1347383400 3900 0 FF
T Star Trek - Deep Space Nine
S Die Erpressung
D Genre: Science Fiction|Als Keiko von einer Reise nach Bajor zurückkehrt, muss O'Brien feststellen, dass ein Pah-Geist von ihr Besitz ergriffen hat. Der Geist will, dass die Wurmloch-Wesen vernichtet werden, damit er wieder in den Himmlischen Tempel einziehen kann. O'Brien soll alle technischen Vorkehrungen dafür treffen - sonst muss Keiko sterben...|Land: USA|Schauspieler: Avery Brooks (Benjamin Sisko) - Rene Auberjonois (Odo) - Cirroc Lofton (Jake Sisko) - Alexander Siddig (Dr. Bashir) - Colm Meaney (Chief O´Brien) - Armin Shimerman (Quark) - Michael Dorn (Lt. Worf)|Jahr: 1996|Regisseur: Jonathan West|Stereo: Ja|Folge Nr.: 5|Genre: Science Fiction
X 2 03 deu Stereo
X 2 05 deu
F 25
P 50
L 99
H
The messages from libavcodec are errors. The enlargement of the logos doesn’t change the result. If I don't have the errors I get proper logos. My suspicion is that this error also handicaps markad to find logos.
Updated by Joe_D over 11 years ago
- % Done changed from 80 to 90
To be able to investigate this further i would need an recording which doesn't provide logos...
Updated by satellit over 11 years ago
I cannot reproduce it. I will come back if it happens again. Thanks
Updated by satellit over 11 years ago
- File TELE_5_HD_Logo.zip TELE_5_HD_Logo.zip added
I found a good example. I get many of these errors:
[h264 @ 0x8267f50]number of reference frames exceeds max (probably corrupt input), discarding one
Last message repeated 1 times
I attached the obtained logos. I started the extraction with this command:
markad --extractlogo=1,430,130 - /media/reel/recordings/Star_Trek_-_Deep_Space_Nine/2012-09-16.20.13.12-0.rec
Generally speaking, markad doesn't work well on Tele 5 HD. Here is the record:
https://rapidshare.com/files/411671309/00001.zip
https://rapidshare.com/files/3913528073/00002.zip
Updated by Joe_D over 11 years ago
Thanks for the recording. For those recordings you dont need extracted logos. I added vertical border detection to the latest GIT.
Can you test it against this recording and report the results?
Updated by satellit over 11 years ago
Thanks for the feedback. I'll try the latest git and let you know.
Updated by satellit over 11 years ago
I tried the 1.4 on Star_Trek_-_Deep_Space_Nine/2012-09-16.20.13.12-0.rec and on a second Star Trek movie. Be aware that I had some errors during build. Maybe is has some effects on the ad detection.
Anyway, the results are the following:
1) the start marks are perfect. Great work!
2) one stop is ok (not perfect) for Star_Trek_-_Deep_Space_Nine/2012-09-16.20.13.12-0.rec. The stop of the second movie is completely off.
3) no advertisements are detected at all. I checked the length of the advertisements which are longer than 5 minutes.
Here are the marks for Star_Trek_-_Deep_Space_Nine/2012-09-16.20.13.12-0.rec, the movie which I uploaded:
marks:
0:01:57.04 detected start of vert. borders (2928)*
1:01:57.22 assuming stop (92946)
marks_0.1.3 (this are the marks with the 0.1.3 version):
0:00:00.01 aspectratio start with 4:3 (0)*
0:01:58.11 detected logo start (2960)*
1:11:54.14 stop of recording (107863)
Updated by Joe_D over 11 years ago
- Status changed from Feedback to Closed
- % Done changed from 90 to 100