I'm trying to tune comskip for channel 99 (Go!) in australia. I'm having trouble understanding the log file of why comskip is combining cblocks. Specifically the blocks are defined as;
Printing AR cblock list before cleaning
-----------------------------------------
Block: 0 Start: 1 End: 786 AR_R: 0.98 Length: 0:00:31.39, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 1 Start: 787 End: 995 AR_R: 0.00 Length: 0:00:08.32, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 2 Start: 996 End: 7386 AR_R: 0.94 Length: 0:04:15.59, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 3 Start: 7387 End: 10893 AR_R: 1.23 Length: 0:02:20.24, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 4 Start: 10894 End: 20944 AR_R: 0.94 Length: 0:06:42.00, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 5 Start: 20945 End: 26211 AR_R: 1.23 Length: 0:03:30.63, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 6 Start: 26212 End: 38581 AR_R: 0.94 Length: 0:08:14.75, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 7 Start: 38582 End: 42091 AR_R: 1.23 Length: 0:02:20.36, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 8 Start: 42092 End: 46235 AR_R: 0.94 Length: 0:02:45.71, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 9 Start: 46236 End: 49839 AR_R: 1.23 Length: 0:02:24.12, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 10 Start: 49840 End: 49975 AR_R: 1.47 Length: 0:00:05.40, [ 704x 576] minX= 1, minY= 54, maxX=704, maxY=526
Block: 11 Start: 49976 End: 50599 AR_R: 1.30 Length: 0:00:24.92, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
However when it gets to AR block sorting the log states that specific block are joined together based upon the logo and the AR being the same.
After Sorting - 98
--------------
Aspect Ratio 0.94 found on 33742 frames totalling 67.0%
Aspect Ratio 1.23 found on 16512 frames totalling 99.7%
Aspect Ratio 1.49 found on 136 frames totalling 100.0%
Joining AR blocks 9 and 10 because both have logo
Joining AR blocks 9 and 10 because both have an AR ratio of 1.23
Joining AR blocks 4 and 5 because both have logo
Joining AR blocks 4 and 5 because both have an AR ratio of 0.94
Joining AR blocks 2 and 3 because both have logo
Joining AR blocks 2 and 3 because both have an AR ratio of 0.94
Joining AR blocks 0 and 2 because they have a dummy cblock inbetween
Quite obviously the AR of these blocks that are being joined are not the same at all. They are all different. Is the logo detection have priority over the AR?? The sentences that the AR are the same is obviously incorrect so I'm trying to understand what comskip is doing to join the blocks.
Thanks
Comskip cblock combination rules
Re: Comskip cblock combination rules
You should not look at the block list before cleaning but to the block list after cleaning as a priority histogram is used to find usefull buckets of AR.
The rule to join AR blocks when both have logo is used for shows that change aspect ratio but are still show.
The rule to join AR blocks when both have logo is used for shows that change aspect ratio but are still show.
Re: Comskip cblock combination rules
If you need more ar buckets you can change the bucket size by adding this setting to comskip.ini
ar_delta=0.04
ar_delta=0.04
Re: Comskip cblock combination rules
I managed to improve my output after many hours of retrying and using virtual dub etc to look at times. I have managed to get all the ads out except for 1 that I really can't find an alternative for after many hours. I've set logo_shrink=8 after noticing that the logo overrun the end of the show and also didn;t turn up immediately at the start of the show. This improved the output immensely. However, I have a set of ads that are for the broadcaster itself that are presenting a problem.
These last for 10 seconds and have the log present in them for start to finish. at the end they cut to a real commercial without any black frames at all.
The sequence of events is that the preceding show has black frames that are detected, then the broadcasters ad cuts in with the logo and the logo is detected and comskip appears to think its part of the previous show. At some point it reaches the early part of the actual ad and decides that its a commercial. I've tried almost everything I can to resolve this it seems that comskip adds the blocks together no matter what.
In the printout below block 2 prior to cleaning represents the exact break that I need to remove. Its perfectly detected via the AR ratio changes. However during the cleaning process blocks 1 and 2 are joined and then the end result is that the cut now starts at 11335.
Any assistence would be greatly appreciated.
Set Logo Quality = 0.82500
The last ar cblock wasn't closed. Now closing.
Printing AR cblock list before cleaning
-----------------------------------------
Block: 0 Start: 1 End: 35 AR_R: 1.24 Length: 0:00:01.36, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 1 Start: 36 End: 10086 AR_R: 0.94 Length: 0:06:42.00, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 2 Start: 10087 End: 15353 AR_R: 1.23 Length: 0:03:30.63, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 3 Start: 15354 End: 17125 AR_R: 0.94 Length: 0:01:10.84, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
After Sorting - 81
--------------
Aspect Ratio 0.94 found on 8194 frames totalling 51.4%
Aspect Ratio 1.23 found on 3847 frames totalling 75.5%
Aspect Ratio 0.96 found on 2260 frames totalling 89.6%
Aspect Ratio 0.95 found on 1627 frames totalling 99.8%
Aspect Ratio 0.92 found on 12 frames totalling 99.9%
Aspect Ratio 0.73 found on 4 frames totalling 99.9%
Aspect Ratio 0.93 found on 4 frames totalling 100.0%
Aspect Ratio 0.97 found on 4 frames totalling 100.0%
Aspect Ratio 1.34 found on 1 frames totalling 100.0%
Aspect Ratio 1.39 found on 1 frames totalling 100.0%
After Sorting - 73
--------------
Aspect Ratio 0.94 found on 11823 frames totalling 69.0%
Aspect Ratio 1.23 found on 5302 frames totalling 100.0%
Joining AR blocks 1 and 2 because both have logo
Joining AR blocks 1 and 2 because both have an AR ratio of 0.94
Printing AR cblock list
-----------------------------------------
Block: 0 Start: 1 End: 35 AR_R: 1.23 Length: 0:00:01.36, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 1 Start: 36 End: 17125 AR_R: 0.94 Length: 0:11:23.55, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Closed caption transcript
--------------------
0) S: 1 E: 17125 L: 0
Distribution of Black Frame cutting: 4 positive and 0 negative, ratio is 9.9900
Confidence of Black Frame cutting: 1 out of 2 are strict, too low
Distribution of Scene Change cutting: 5 positive and 0 negative, ratio is 9.9900
Confidence of Scene Change cutting: 3 out of 4 are strict
Distribution of Uniform Frame cutting: 1 positive and 0 negative, ratio is 9.9900
Confidence of Uniform Frame cutting: 1 out of 1 are strict, too low
Distribution of Volume cutting: 3 positive and 0 negative, ratio is 9.9900
Confidence of Volume cutting: 3 out of 3 are strict
Black Frame List
---------------------------
Black Frame Count = 112
nr frame bright uniform volume cause dimcount bright
0 26 21 359 5 b 0 0
1 27 16 0 5 a b 0 0
2 28 16 3 5 b 0 0
3 29 16 0 5 b 0 0
4 30 16 0 5 b 0 0
5 31 17 119 5 b 0 0
6 32 20 583 5 b 0 0
7 33 24 2513 5 b 3 0
8 34 28 4219 5 b 720 0
-----------------------------
9 10069 19 241 5 b 26 1
10 10070 19 101 5 b 17 0
11 10071 18 83 5 b 16 0
12 10072 17 42 5 b 14 0
13 10073 17 23 5 b 14 0
14 10074 16 18 5 b 12 0
15 10075 16 15 5 b 3 0
16 10076 16 12 5 b 0 0
17 10077 16 10 5 b 0 0
18 10078 16 7 5 b 0 0
19 10079 16 5 5 b 0 0
20 10080 16 3 5 b 0 0
21 10081 16 0 5 b 0 0
22 10082 16 0 5 b 0 0
23 10083 16 0 5 b 0 0
24 10084 18 215 5 b 0 0
25 10085 27 3008 5 b 18 0
-----------------------------
26 11323 91 26945 9 v 4443 177
27 11324 89 26728 9 v 4449 178
28 11325 87 26582 9 v 4471 177
29 11326 86 27088 9 v 4606 176
30 11327 85 26866 9 v 4558 176
31 11328 83 26415 9 v 4663 179
32 11329 83 26523 9 v 4563 175
33 11330 82 25888 9 v 4535 177
34 11331 83 25384 9 v 4592 177
35 11332 82 24908 9 v 4734 175
36 11333 81 24670 9 v 4852 178
37 11334 89 26853 9 v 2952 169
38 11335 89 26876 9 v 2950 170
39 11336 89 26948 9 v 2970 168
40 11337 89 27013 9 v 2945 167
41 11338 89 27014 9 v 2954 167
42 11339 89 26901 9 v 2979 163
43 11340 89 26898 9 v 2986 165
44 11341 89 26931 9 v 2983 165
45 11342 89 26843 9 v 3006 163
46 11343 89 26815 9 v 3023 165
47 11344 88 26877 9 v 3015 167
48 11345 89 26868 9 v 3013 167
49 11346 89 26942 9 v 3013 167
50 11347 89 26993 9 v 3003 168
-----------------------------
51 12834 16 67 9 s 2 8
-----------------------------
52 12837 60 20047 350 s 6539 161
-----------------------------
53 13575 204 10999 0 v 19 171
54 13576 204 10939 0 v 22 172
55 13577 204 10947 0 v 15 170
56 13578 204 10903 0 v 16 172
57 13579 204 10883 0 v 17 173
58 13580 204 10858 0 v 9 172
59 13581 204 10922 0 v 9 170
60 13582 204 10872 0 v 11 172
61 13583 204 10781 0 v 14 172
62 13584 16 0 0 v b 0 0
63 13585 16 0 0 v b 0 0
64 13586 15 0 0 v b 0 0
65 13587 144 26649 0 v 192 175
66 13588 144 26565 0 v 211 171
67 13589 143 26938 0 v 212 172
68 13590 143 26924 0 v 207 173
69 13591 143 27168 0 v 223 174
70 13592 142 27321 0 v 276 172
71 13593 142 27566 0 v 253 173
72 13594 142 27622 0 v 224 172
73 13595 141 27526 0 v 232 173
74 13596 140 27877 0 v 266 171
75 13597 140 27979 0 v 269 171
76 13598 140 28159 0 v 245 173
77 13599 139 28331 0 v 307 171
78 13600 139 28396 2 v 262 174
79 13601 138 28746 5 v 273 172
-----------------------------
80 14325 218 13309 7 v 3 167
81 14326 218 13318 7 v 3 172
82 14327 218 13302 7 v 3 169
83 14328 218 13296 7 v 4 170
84 14329 218 13292 7 v 4 169
85 14330 218 13284 7 v 4 171
86 14331 218 13302 7 v 4 170
87 14332 218 13290 7 v 4 170
88 14333 218 13290 7 v 4 168
89 14334 35 16263 7 vs 919 170
90 14335 35 16269 7 v 929 171
91 14336 35 16276 7 v 925 165
92 14337 36 17166 7 v 774 168
93 14338 36 17730 7 v 682 165
94 14339 34 16739 7 v 808 166
95 14340 35 17416 7 v 1078 166
96 14341 35 17379 7 v 1090 166
97 14342 35 17439 7 v 1055 166
-----------------------------
98 15084 96 23299 350 s 2737 153
-----------------------------
99 15240 145 28696 350 s 0 73
-----------------------------
100 15333 21 257 5 b 0 0
101 15334 17 100 5 b 0 0
102 15335 17 0 5 b 0 0
103 15336 16 0 5 b 0 0
104 15337 16 0 5 b 0 0
105 15338 16 0 5 b 0 0
106 15339 16 0 5 b 0 0
107 15340 16 0 5 b 0 0
108 15341 17 76 5 b 0 0
109 15342 18 523 5 b 0 0
110 15343 21 2229 5 b 151 0
-----------------------------
111 17116 73 24846 515 a 3221 155
Logos detected on the following frames
--------------------------------------
Logo start - 325 end - 10350 length - 0:06:41.00 before:11.8 s after:39.4 s
Logo start - 15925 end - 16900 length - 0:00:39.00 before:23.5 s after:8.6 s
Combining CC Blocks...
CC's detected on the following frames - 1 total blocks
--------------------------------------
0 - CC start - 0 end - 17125 type - NONE length - 0:11:25.00
Caption sums
---------------------------
Pop on captions: 0: 0.00 - 0:00:00.00
Roll up captions: 0: 0.00 - 0:00:00.00
Paint on captions: 0: 0.00 - 0:00:00.00
No captions: 17126:99.85 - 0:11:25.03
The NONE type of closed captions were determined to be the most common.
Start - 1 End - 29 CCF - 1 CCL - 0 Type - NONE
Start - 30 End - 10076 CCF - 1 CCL - 0 Type - NONE
Start - 10077 End - 11334 CCF - 1 CCL - 0 Type - NONE
Start - 11335 End - 12835 CCF - 1 CCL - 0 Type - NONE
Start - 12836 End - 13584 CCF - 1 CCL - 0 Type - NONE
Start - 13585 End - 14333 CCF - 1 CCL - 0 Type - NONE
Start - 14334 End - 15083 CCF - 1 CCL - 0 Type - NONE
Start - 15084 End - 15239 CCF - 1 CCL - 0 Type - NONE
Start - 15240 End - 15337 CCF - 1 CCL - 0 Type - NONE
Start - 15338 End - 17115 CCF - 1 CCL - 0 Type - NONE
Start - 17116 End - 17152 CCF - 1 CCL - 501 Type - NONE
Joining blocks 1 and 2 at frame 10076 because they both have a logo.
Block 000 schange_rate - 2.21 average - 0.08
Block 001 schange_rate - 0.04 average - 0.08
Block 002 schange_rate - 0.08 average - 0.08
Block 003 schange_rate - 0.33 average - 0.08
Block 004 schange_rate - 0.10 average - 0.08
Block 005 schange_rate - 0.33 average - 0.08
Block 006 schange_rate - 0.10 average - 0.08
Block 007 schange_rate - 0.88 average - 0.08
Block 008 schange_rate - 0.02 average - 0.08
Block 009 schange_rate - 0.00 average - 0.08
Fuzzy scoring of the blocks
---------------------------
Block 0 has no logo.
Block 0 score: Before - 1.00 After - 2.00
Block 0 has more scene change than average.
Block 0 score: Before - 2.00 After - 4.00
Block 0 AR (1.23) is different from dominant AR(0.94).
Block 0 score: Before - 4.00 After - 8.00
Block 1 has logo.
Block 1 score: Before - 1.00 After - 0.01
Block 1 has twice excess length.
Block 1 score: Before - 0.01 After - 0.00
Block 2 has strict standard length for a commercial.
Block 2 score: Before - 1.00 After - 3.00
Block 2 has no logo.
Block 2 score: Before - 3.00 After - 6.00
Block 3 has strict standard length for a commercial.
Block 3 score: Before - 1.00 After - 3.00
Block 3 has no logo.
Block 3 score: Before - 3.00 After - 6.00
Block 3 has more scene change than average.
Block 3 score: Before - 6.00 After - 12.00
Block 4 has strict standard length for a commercial.
Block 4 score: Before - 1.00 After - 3.00
Block 4 has no logo.
Block 4 score: Before - 3.00 After - 6.00
Block 4 is much louder than average.
Block 4 score: Before - 6.00 After - 12.00
Block 4 has less silence than average.
Block 4 score: Before - 12.00 After - 24.00
Block 5 has strict standard length for a commercial.
Block 5 score: Before - 1.00 After - 3.00
Combining Blocks 5 thru 7 result in non-strict standard commercial length of 40.04 with a tolerance of 0.360000.
Block 5 score: Before - 3.00 After - 3.63 Combined count - 1
Block 6 score: Before - 1.00 After - 1.21 Combined count - 1
Block 7 score: Before - 1.00 After - 1.21 Combined count - 1
Block 5 has no logo.
Block 5 score: Before - 3.63 After - 7.25
Block 5 is much louder than average.
Block 5 score: Before - 7.25 After - 14.50
Block 5 has less silence than average.
Block 5 score: Before - 14.50 After - 29.00
Block 5 has more scene change than average.
Block 5 score: Before - 29.00 After - 58.00
Combining Blocks 6 thru 7 result in strict standard commercial length of 10.08 with a tolerance of 0.360000.
Block 6 score: Before - 1.21 After - 1.81 Combined count - 2
Block 7 score: Before - 1.21 After - 1.81 Combined count - 2
Block 6 has no logo.
Block 6 score: Before - 1.81 After - 3.63
Combining Blocks 6 thru 7 result in strict standard commercial length of 10.08 with a tolerance of 0.360000.
Block 7 score: Before - 1.81 After - 2.72 Combined count - 3
Block 6 score: Before - 3.63 After - 5.44 Combined count - 3
Block 7 has no logo.
Block 7 score: Before - 2.72 After - 5.44
Block 7 has more scene change than average.
Block 7 score: Before - 5.44 After - 10.88
Block 8 has logo.
Block 8 score: Before - 1.00 After - 0.01
Block 9 has no logo.
Block 9 score: Before - 1.00 After - 2.00
---------------------
Initial Commercial List
---------------------
0) 1 29 0:00:01.12
1) 11335 15337 0:02:40.08
2) 17116 17152 0:00:01.43
No change
Logo fraction: 0.6599
Maximum volume found: 3405
Average volume: 834
Sound threshold: 700
Silence threshold: 100
Minimum volume found: 0
Average frames with silence: 55
Black threshold: 28
Minimum brightness found: 16
Minimum bright pixels found: 0
Minimum dim level found: 0
Average brightness: 443
Uniformity level: 13200
Average non uniformity: 199
Maximum gap between logo's: 223
Suggested logo_threshold: 0.8250
Suggested shrink_logo: 8
Max commercial size found: 60
Dominant aspect ratio: 0.9400
Score threshold: 1.0500
Framerate: 25.00
Total commercial length: 0:02:41.20
Block list after weighing
----------------------------------------------------
# sbf bs be fs fe sc len scr cmb ar cut bri logo vol sil corr stdev cc
0:-- 4 0 4 1 29 8.00 1.120 2.21 0 1.23 A L a b 9068+ 0.00 416- 27- 0.000 906 NONE
1:++ 13 5 12 30 11334 0.00 452.160 0.04 0 0.94 EL v 273- 0.89 746- 49- 0.000 27 NONE
2:-- 15 13 2 11335 12835 6.00 60.000 0.08 0 0.94 L S s 598+ 0.00 1050+ 64+ 0.000 59 NONE
3:-- 3 2 1 12836 13584 12.00 29.920 0.33 0 0.94 L S v b 822+ 0.00 1025+ 61+ 0.000 82 NONE
4:-- 2 2 0 13585 14333 24.00 29.920 0.10 0 0.94 L S vs 753+ 0.00 1221+ 84+ 0.000 75 NONE
5:-- 1 1 0 14334 15083 58.00 29.960 0.33 1 0.94 LC S s 1785+ 0.00 1239+ 87+ 0.000 178 NONE
6:-- 1 1 0 15084 15239 5.44 6.200 0.10 3 0.94 LC s 1384+ 0.00 1002+ 70+ 0.000 138 NONE
7:-- 6 1 5 15240 15337 10.88 3.880 0.88 3 0.94 LC b 2316+ 0.00 931+ 63+ 0.000 231 NONE
8:++ 6 6 0 15338 17115 0.01 71.080 0.02 0 0.94 L a 176- 0.55 797- 50- 0.000 17 NONE
9:-- 1 1 0 17116 17152 2.00 1.440 0.00 0 0.94 L b 1972+ 0.00 160- 10- 0.000 197 NONE
These last for 10 seconds and have the log present in them for start to finish. at the end they cut to a real commercial without any black frames at all.
The sequence of events is that the preceding show has black frames that are detected, then the broadcasters ad cuts in with the logo and the logo is detected and comskip appears to think its part of the previous show. At some point it reaches the early part of the actual ad and decides that its a commercial. I've tried almost everything I can to resolve this it seems that comskip adds the blocks together no matter what.
In the printout below block 2 prior to cleaning represents the exact break that I need to remove. Its perfectly detected via the AR ratio changes. However during the cleaning process blocks 1 and 2 are joined and then the end result is that the cut now starts at 11335.
Any assistence would be greatly appreciated.
Set Logo Quality = 0.82500
The last ar cblock wasn't closed. Now closing.
Printing AR cblock list before cleaning
-----------------------------------------
Block: 0 Start: 1 End: 35 AR_R: 1.24 Length: 0:00:01.36, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 1 Start: 36 End: 10086 AR_R: 0.94 Length: 0:06:42.00, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Block: 2 Start: 10087 End: 15353 AR_R: 1.23 Length: 0:03:30.63, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 3 Start: 15354 End: 17125 AR_R: 0.94 Length: 0:01:10.84, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
After Sorting - 81
--------------
Aspect Ratio 0.94 found on 8194 frames totalling 51.4%
Aspect Ratio 1.23 found on 3847 frames totalling 75.5%
Aspect Ratio 0.96 found on 2260 frames totalling 89.6%
Aspect Ratio 0.95 found on 1627 frames totalling 99.8%
Aspect Ratio 0.92 found on 12 frames totalling 99.9%
Aspect Ratio 0.73 found on 4 frames totalling 99.9%
Aspect Ratio 0.93 found on 4 frames totalling 100.0%
Aspect Ratio 0.97 found on 4 frames totalling 100.0%
Aspect Ratio 1.34 found on 1 frames totalling 100.0%
Aspect Ratio 1.39 found on 1 frames totalling 100.0%
After Sorting - 73
--------------
Aspect Ratio 0.94 found on 11823 frames totalling 69.0%
Aspect Ratio 1.23 found on 5302 frames totalling 100.0%
Joining AR blocks 1 and 2 because both have logo
Joining AR blocks 1 and 2 because both have an AR ratio of 0.94
Printing AR cblock list
-----------------------------------------
Block: 0 Start: 1 End: 35 AR_R: 1.23 Length: 0:00:01.36, [ 704x 576] minX= 1, minY= 1, maxX=704, maxY=576
Block: 1 Start: 36 End: 17125 AR_R: 0.94 Length: 0:11:23.55, [ 704x 576] minX= 82, minY= 1, maxX=622, maxY=576
Closed caption transcript
--------------------
0) S: 1 E: 17125 L: 0
Distribution of Black Frame cutting: 4 positive and 0 negative, ratio is 9.9900
Confidence of Black Frame cutting: 1 out of 2 are strict, too low
Distribution of Scene Change cutting: 5 positive and 0 negative, ratio is 9.9900
Confidence of Scene Change cutting: 3 out of 4 are strict
Distribution of Uniform Frame cutting: 1 positive and 0 negative, ratio is 9.9900
Confidence of Uniform Frame cutting: 1 out of 1 are strict, too low
Distribution of Volume cutting: 3 positive and 0 negative, ratio is 9.9900
Confidence of Volume cutting: 3 out of 3 are strict
Black Frame List
---------------------------
Black Frame Count = 112
nr frame bright uniform volume cause dimcount bright
0 26 21 359 5 b 0 0
1 27 16 0 5 a b 0 0
2 28 16 3 5 b 0 0
3 29 16 0 5 b 0 0
4 30 16 0 5 b 0 0
5 31 17 119 5 b 0 0
6 32 20 583 5 b 0 0
7 33 24 2513 5 b 3 0
8 34 28 4219 5 b 720 0
-----------------------------
9 10069 19 241 5 b 26 1
10 10070 19 101 5 b 17 0
11 10071 18 83 5 b 16 0
12 10072 17 42 5 b 14 0
13 10073 17 23 5 b 14 0
14 10074 16 18 5 b 12 0
15 10075 16 15 5 b 3 0
16 10076 16 12 5 b 0 0
17 10077 16 10 5 b 0 0
18 10078 16 7 5 b 0 0
19 10079 16 5 5 b 0 0
20 10080 16 3 5 b 0 0
21 10081 16 0 5 b 0 0
22 10082 16 0 5 b 0 0
23 10083 16 0 5 b 0 0
24 10084 18 215 5 b 0 0
25 10085 27 3008 5 b 18 0
-----------------------------
26 11323 91 26945 9 v 4443 177
27 11324 89 26728 9 v 4449 178
28 11325 87 26582 9 v 4471 177
29 11326 86 27088 9 v 4606 176
30 11327 85 26866 9 v 4558 176
31 11328 83 26415 9 v 4663 179
32 11329 83 26523 9 v 4563 175
33 11330 82 25888 9 v 4535 177
34 11331 83 25384 9 v 4592 177
35 11332 82 24908 9 v 4734 175
36 11333 81 24670 9 v 4852 178
37 11334 89 26853 9 v 2952 169
38 11335 89 26876 9 v 2950 170
39 11336 89 26948 9 v 2970 168
40 11337 89 27013 9 v 2945 167
41 11338 89 27014 9 v 2954 167
42 11339 89 26901 9 v 2979 163
43 11340 89 26898 9 v 2986 165
44 11341 89 26931 9 v 2983 165
45 11342 89 26843 9 v 3006 163
46 11343 89 26815 9 v 3023 165
47 11344 88 26877 9 v 3015 167
48 11345 89 26868 9 v 3013 167
49 11346 89 26942 9 v 3013 167
50 11347 89 26993 9 v 3003 168
-----------------------------
51 12834 16 67 9 s 2 8
-----------------------------
52 12837 60 20047 350 s 6539 161
-----------------------------
53 13575 204 10999 0 v 19 171
54 13576 204 10939 0 v 22 172
55 13577 204 10947 0 v 15 170
56 13578 204 10903 0 v 16 172
57 13579 204 10883 0 v 17 173
58 13580 204 10858 0 v 9 172
59 13581 204 10922 0 v 9 170
60 13582 204 10872 0 v 11 172
61 13583 204 10781 0 v 14 172
62 13584 16 0 0 v b 0 0
63 13585 16 0 0 v b 0 0
64 13586 15 0 0 v b 0 0
65 13587 144 26649 0 v 192 175
66 13588 144 26565 0 v 211 171
67 13589 143 26938 0 v 212 172
68 13590 143 26924 0 v 207 173
69 13591 143 27168 0 v 223 174
70 13592 142 27321 0 v 276 172
71 13593 142 27566 0 v 253 173
72 13594 142 27622 0 v 224 172
73 13595 141 27526 0 v 232 173
74 13596 140 27877 0 v 266 171
75 13597 140 27979 0 v 269 171
76 13598 140 28159 0 v 245 173
77 13599 139 28331 0 v 307 171
78 13600 139 28396 2 v 262 174
79 13601 138 28746 5 v 273 172
-----------------------------
80 14325 218 13309 7 v 3 167
81 14326 218 13318 7 v 3 172
82 14327 218 13302 7 v 3 169
83 14328 218 13296 7 v 4 170
84 14329 218 13292 7 v 4 169
85 14330 218 13284 7 v 4 171
86 14331 218 13302 7 v 4 170
87 14332 218 13290 7 v 4 170
88 14333 218 13290 7 v 4 168
89 14334 35 16263 7 vs 919 170
90 14335 35 16269 7 v 929 171
91 14336 35 16276 7 v 925 165
92 14337 36 17166 7 v 774 168
93 14338 36 17730 7 v 682 165
94 14339 34 16739 7 v 808 166
95 14340 35 17416 7 v 1078 166
96 14341 35 17379 7 v 1090 166
97 14342 35 17439 7 v 1055 166
-----------------------------
98 15084 96 23299 350 s 2737 153
-----------------------------
99 15240 145 28696 350 s 0 73
-----------------------------
100 15333 21 257 5 b 0 0
101 15334 17 100 5 b 0 0
102 15335 17 0 5 b 0 0
103 15336 16 0 5 b 0 0
104 15337 16 0 5 b 0 0
105 15338 16 0 5 b 0 0
106 15339 16 0 5 b 0 0
107 15340 16 0 5 b 0 0
108 15341 17 76 5 b 0 0
109 15342 18 523 5 b 0 0
110 15343 21 2229 5 b 151 0
-----------------------------
111 17116 73 24846 515 a 3221 155
Logos detected on the following frames
--------------------------------------
Logo start - 325 end - 10350 length - 0:06:41.00 before:11.8 s after:39.4 s
Logo start - 15925 end - 16900 length - 0:00:39.00 before:23.5 s after:8.6 s
Combining CC Blocks...
CC's detected on the following frames - 1 total blocks
--------------------------------------
0 - CC start - 0 end - 17125 type - NONE length - 0:11:25.00
Caption sums
---------------------------
Pop on captions: 0: 0.00 - 0:00:00.00
Roll up captions: 0: 0.00 - 0:00:00.00
Paint on captions: 0: 0.00 - 0:00:00.00
No captions: 17126:99.85 - 0:11:25.03
The NONE type of closed captions were determined to be the most common.
Start - 1 End - 29 CCF - 1 CCL - 0 Type - NONE
Start - 30 End - 10076 CCF - 1 CCL - 0 Type - NONE
Start - 10077 End - 11334 CCF - 1 CCL - 0 Type - NONE
Start - 11335 End - 12835 CCF - 1 CCL - 0 Type - NONE
Start - 12836 End - 13584 CCF - 1 CCL - 0 Type - NONE
Start - 13585 End - 14333 CCF - 1 CCL - 0 Type - NONE
Start - 14334 End - 15083 CCF - 1 CCL - 0 Type - NONE
Start - 15084 End - 15239 CCF - 1 CCL - 0 Type - NONE
Start - 15240 End - 15337 CCF - 1 CCL - 0 Type - NONE
Start - 15338 End - 17115 CCF - 1 CCL - 0 Type - NONE
Start - 17116 End - 17152 CCF - 1 CCL - 501 Type - NONE
Joining blocks 1 and 2 at frame 10076 because they both have a logo.
Block 000 schange_rate - 2.21 average - 0.08
Block 001 schange_rate - 0.04 average - 0.08
Block 002 schange_rate - 0.08 average - 0.08
Block 003 schange_rate - 0.33 average - 0.08
Block 004 schange_rate - 0.10 average - 0.08
Block 005 schange_rate - 0.33 average - 0.08
Block 006 schange_rate - 0.10 average - 0.08
Block 007 schange_rate - 0.88 average - 0.08
Block 008 schange_rate - 0.02 average - 0.08
Block 009 schange_rate - 0.00 average - 0.08
Fuzzy scoring of the blocks
---------------------------
Block 0 has no logo.
Block 0 score: Before - 1.00 After - 2.00
Block 0 has more scene change than average.
Block 0 score: Before - 2.00 After - 4.00
Block 0 AR (1.23) is different from dominant AR(0.94).
Block 0 score: Before - 4.00 After - 8.00
Block 1 has logo.
Block 1 score: Before - 1.00 After - 0.01
Block 1 has twice excess length.
Block 1 score: Before - 0.01 After - 0.00
Block 2 has strict standard length for a commercial.
Block 2 score: Before - 1.00 After - 3.00
Block 2 has no logo.
Block 2 score: Before - 3.00 After - 6.00
Block 3 has strict standard length for a commercial.
Block 3 score: Before - 1.00 After - 3.00
Block 3 has no logo.
Block 3 score: Before - 3.00 After - 6.00
Block 3 has more scene change than average.
Block 3 score: Before - 6.00 After - 12.00
Block 4 has strict standard length for a commercial.
Block 4 score: Before - 1.00 After - 3.00
Block 4 has no logo.
Block 4 score: Before - 3.00 After - 6.00
Block 4 is much louder than average.
Block 4 score: Before - 6.00 After - 12.00
Block 4 has less silence than average.
Block 4 score: Before - 12.00 After - 24.00
Block 5 has strict standard length for a commercial.
Block 5 score: Before - 1.00 After - 3.00
Combining Blocks 5 thru 7 result in non-strict standard commercial length of 40.04 with a tolerance of 0.360000.
Block 5 score: Before - 3.00 After - 3.63 Combined count - 1
Block 6 score: Before - 1.00 After - 1.21 Combined count - 1
Block 7 score: Before - 1.00 After - 1.21 Combined count - 1
Block 5 has no logo.
Block 5 score: Before - 3.63 After - 7.25
Block 5 is much louder than average.
Block 5 score: Before - 7.25 After - 14.50
Block 5 has less silence than average.
Block 5 score: Before - 14.50 After - 29.00
Block 5 has more scene change than average.
Block 5 score: Before - 29.00 After - 58.00
Combining Blocks 6 thru 7 result in strict standard commercial length of 10.08 with a tolerance of 0.360000.
Block 6 score: Before - 1.21 After - 1.81 Combined count - 2
Block 7 score: Before - 1.21 After - 1.81 Combined count - 2
Block 6 has no logo.
Block 6 score: Before - 1.81 After - 3.63
Combining Blocks 6 thru 7 result in strict standard commercial length of 10.08 with a tolerance of 0.360000.
Block 7 score: Before - 1.81 After - 2.72 Combined count - 3
Block 6 score: Before - 3.63 After - 5.44 Combined count - 3
Block 7 has no logo.
Block 7 score: Before - 2.72 After - 5.44
Block 7 has more scene change than average.
Block 7 score: Before - 5.44 After - 10.88
Block 8 has logo.
Block 8 score: Before - 1.00 After - 0.01
Block 9 has no logo.
Block 9 score: Before - 1.00 After - 2.00
---------------------
Initial Commercial List
---------------------
0) 1 29 0:00:01.12
1) 11335 15337 0:02:40.08
2) 17116 17152 0:00:01.43
No change
Logo fraction: 0.6599
Maximum volume found: 3405
Average volume: 834
Sound threshold: 700
Silence threshold: 100
Minimum volume found: 0
Average frames with silence: 55
Black threshold: 28
Minimum brightness found: 16
Minimum bright pixels found: 0
Minimum dim level found: 0
Average brightness: 443
Uniformity level: 13200
Average non uniformity: 199
Maximum gap between logo's: 223
Suggested logo_threshold: 0.8250
Suggested shrink_logo: 8
Max commercial size found: 60
Dominant aspect ratio: 0.9400
Score threshold: 1.0500
Framerate: 25.00
Total commercial length: 0:02:41.20
Block list after weighing
----------------------------------------------------
# sbf bs be fs fe sc len scr cmb ar cut bri logo vol sil corr stdev cc
0:-- 4 0 4 1 29 8.00 1.120 2.21 0 1.23 A L a b 9068+ 0.00 416- 27- 0.000 906 NONE
1:++ 13 5 12 30 11334 0.00 452.160 0.04 0 0.94 EL v 273- 0.89 746- 49- 0.000 27 NONE
2:-- 15 13 2 11335 12835 6.00 60.000 0.08 0 0.94 L S s 598+ 0.00 1050+ 64+ 0.000 59 NONE
3:-- 3 2 1 12836 13584 12.00 29.920 0.33 0 0.94 L S v b 822+ 0.00 1025+ 61+ 0.000 82 NONE
4:-- 2 2 0 13585 14333 24.00 29.920 0.10 0 0.94 L S vs 753+ 0.00 1221+ 84+ 0.000 75 NONE
5:-- 1 1 0 14334 15083 58.00 29.960 0.33 1 0.94 LC S s 1785+ 0.00 1239+ 87+ 0.000 178 NONE
6:-- 1 1 0 15084 15239 5.44 6.200 0.10 3 0.94 LC s 1384+ 0.00 1002+ 70+ 0.000 138 NONE
7:-- 6 1 5 15240 15337 10.88 3.880 0.88 3 0.94 LC b 2316+ 0.00 931+ 63+ 0.000 231 NONE
8:++ 6 6 0 15338 17115 0.01 71.080 0.02 0 0.94 L a 176- 0.55 797- 50- 0.000 17 NONE
9:-- 1 1 0 17116 17152 2.00 1.440 0.00 0 0.94 L b 1972+ 0.00 160- 10- 0.000 197 NONE
Re: Comskip cblock combination rules
That should have said
These last for 10 seconds and have the logo present in them for start to finish
These last for 10 seconds and have the logo present in them for start to finish
Re: Comskip cblock combination rules
Thanks for sending the "how to ask for help" files, makes it much easier to provide some help.
The logo moves into the block you want to be added to the commercial.
One solution is to drastically shrink the logo
shrink_logo=20
another solution is to set
connect_blocks_with_logo=0
logo_present_modifier=1
the ensures the block do not get combined and the small fraction of logo present in the block does not prevent it from being scored as commercial.
The logo moves into the block you want to be added to the commercial.
One solution is to drastically shrink the logo
shrink_logo=20
another solution is to set
connect_blocks_with_logo=0
logo_present_modifier=1
the ensures the block do not get combined and the small fraction of logo present in the block does not prevent it from being scored as commercial.
Re: Comskip cblock combination rules
Its really odd. If I use the debug window then the point in the actual commercial where comskip takes the logo detection from 0.38 to 0 is where the cut point is. However on the frames that are showing a logo detection of 0.38 the csv file shows logo is 0 for that frame. Visually there is no logo in the frame at all but comskip is showing 0.38...no idea why its doing that at all. I've tried adjusting the logo threshold and on my 5 minute sample increasing the threshold to 0.80 fixed the dection issue. I was over joyed...then I retried on the 30minute program and it still had the same logo detection issue...ie. on fromes with no logo its showing 0.38.
Any ideas?
Any ideas?
Re: Comskip cblock combination rules
Logo is averaged over 25 frames as otherwise its not reliable
Look at the yellow line on the debug window
I would need to see what is actually inside the first commercial block that triggers the logo detection to be able to make a suggestion on howto improve.
For your test recording setting
logo_threshold=0.85
produces the desired result but this tweaking of the threshold is NOT very robust
Look at the yellow line on the debug window
I would need to see what is actually inside the first commercial block that triggers the logo detection to be able to make a suggestion on howto improve.
For your test recording setting
logo_threshold=0.85
produces the desired result but this tweaking of the threshold is NOT very robust
Re: Comskip cblock combination rules
totally agree with you on the robustness part. Just can't figure out how to get it to stop doing what it is doing. When I returned to the full 30 minute show I had to increase the threshold to the absolute max of 0.95 for it to work. Works but thats only because of the timing. I sent you the csv etc i'll see if i can chop the file down to something less than a few hundred meg if you have the time to look at it. I got desparate so I actually changed everything that I possibly could and ran it through comskip....only thing that seems to impact the outcome is the threshold and even then it doesn;t work as expected...
Re: Comskip cblock combination rules
I was going to ask.
So when using the debug window and logo is 0.38 for instance how does comskip interpret this? The csv file logs it as no log on these frames but it appears that comskip itself treats them as being logo detected. It appears only when the threshold is upped to the max does the 0.38 level drop to a lesser value that comskip then considers is not a logo.
So when using the debug window and logo is 0.38 for instance how does comskip interpret this? The csv file logs it as no log on these frames but it appears that comskip itself treats them as being logo detected. It appears only when the threshold is upped to the max does the 0.38 level drop to a lesser value that comskip then considers is not a logo.
Re: Comskip cblock combination rules
Don't look at the logo values for individual frames. The algorithm is rather complex
Best is to send me the .csv and .ref of the full recording
Best is to send me the .csv and .ref of the full recording
Re: Comskip cblock combination rules
its in your mail.
interest have over 50 frames without logo and the debug still shows 0.78 on the logo for them all.....
sent you the csv for the cut down version. Taking too long to try all options on the 30 minute show. Playing around with logo detection stuff at present. Its interesting as to how it cuts in the middle of the ad and one side in the debug has logo=0.78 and the point of cut is logo=0. The 30 frames prior to the cut all show logo=0.78.
interest have over 50 frames without logo and the debug still shows 0.78 on the logo for them all.....
sent you the csv for the cut down version. Taking too long to try all options on the 30 minute show. Playing around with logo detection stuff at present. Its interesting as to how it cuts in the middle of the ad and one side in the debug has logo=0.78 and the point of cut is logo=0. The 30 frames prior to the cut all show logo=0.78.
Re: Comskip cblock combination rules
You have set
logo_filter=4
which averages over about 100 frames.
try
logo_filter=0
logo_filter=4
which averages over about 100 frames.
try
logo_filter=0
Re: Comskip cblock combination rules
no difference in the end result. Still drops to a cut point in the middle of the commercial.
Re: Comskip cblock combination rules
i'll up the cut down mpg if it helps. its 200m may take some time...upstream is slow in oz