ShowAnalyzer Beta

[Jere_Jones]

i do get crappy reception, but even so comskip works about 70-80% of the time.

I suspect this may be the problem. Crummy reception affects the audio as much as the video. SA won't mark a frame as a border frame unless it is blank and quiet. Bad reception probably means static in the audio that triggers the "not quiet" flag. There really is no good way around this. Run a show with a command line option of -framedetail and send me the file it creates and the log file. It is likely to be a few megabytes. That will let me see what the problem is.

Jere

[smeghead]

Hi Jere

Just off to work now but could I do the same as reven? I dont get any commercials detected but have pretty good reception from cable in the UK. I've tried lots of tweaks to the settings and detection gets slightly better but still not as good as comskip.

[tieke]

Whereas I get mixed results - on programs captured via my PVR-250, SA is getting basically every single ad break, whereas comskip would spot only three quarters of them. However on programs captured with my dvb-s card, SA picks up only one ad in five, whereas comskip gets them all. To simplify things, I have just cunningly destroyed my dvb card while doing some work around my server, so that's fixed that then

[Jere_Jones]

Hi Jere

Just off to work now but could I do the same as reven? I dont get any commercials detected but have pretty good reception from cable in the UK. I've tried lots of tweaks to the settings and detection gets slightly better but still not as good as comskip.

Absolutely. Have you tried the aggressive segmentation setting? It makes all the difference on a few channels in the UK. Just don't use it with cartoons.

Jere

[Spartan]

Ok, I've had a little time to play around with SA on my Episode 3 of Into the West. Here is a snippet of the logfile (SA running with the default settings):

Code:

Block List
-----------------------------------------------------------------------------------------------------
#    Start    End    Length        Eval of Length        Blanks    EP    SC    AvgSpl    Score
-----------------------------------------------------------------------------------------------------
    Show
0)    0    1464    0:00:48.882    NOT_COMMERCIAL          0    1    0.00    -27.70    1
1)    1465    2984    0:00:50.717    NON_STRICT_COMMERCIAL    23    1    0.00    -27.98    1.5
2)    2985    3064    0:00:02.669    TOO_SHORT               8    0    0.00    -33.08    20
3)    3065    3633    0:00:18.985    NOT_COMMERCIAL          17    1    0.00    -28.35    1
4)    3634    4578    0:00:31.531    NOT_COMMERCIAL          2    1    0.00    -28.11    1
5)    4579    7763    0:01:46.272    TOO_LONG                37    0    0.00    -26.85    2
6)    7764    17114    0:05:12.011    TOO_LONG                26    0    0.00    -28.72    2
7)    17115    35880    0:10:26.158    TOO_LONG                5    0    0.00    -28.58    2
    Commercial
8)    35881    35916    0:00:01.201    TOO_SHORT               3    0    0.00    -35.18    20
9)    35917    35988    0:00:02.402    TOO_SHORT               5    0    0.00    -29.15    20
10)    35989    36125    0:00:04.571    NON_STRICT_COMMERCIAL    10    0    0.00    -30.10    30
11)    36126    36166    0:00:01.368    NON_STRICT_COMMERCIAL    7    0    0.00    -31.58    30
12)    36167    40002    0:02:07.994    TOO_LONG                61    0    0.00    -28.94    2
13)    40003    40064    0:00:02.068    TOO_SHORT               13    0    0.00    -31.21    20
14)    40065    41614    0:00:51.718    NON_STRICT_COMMERCIAL    5    0    0.00    -29.22    30
15)    *41615    42117    0:00:16.783    STRICT_COMMERCIAL        55    0    0.00    -29.75    40
16)    42118    43058    0:00:31.398    STRICT_COMMERCIAL        48    0    0.00    -31.49    40
17)    43059    43506    0:00:14.948    STRICT_COMMERCIAL        2    0    0.00    -30.45    40
18)    43507    43956    0:00:15.015    STRICT_COMMERCIAL        2    0    0.00    -30.86    40
19)    43957    44408    0:00:15.081    STRICT_COMMERCIAL        4    0    0.00    -28.93    40
20)    44409    45308    0:00:30.030    STRICT_COMMERCIAL        6    0    0.00    -29.49    40
21)    45309    45766    0:00:15.281    STRICT_COMMERCIAL        15    0    0.00    -28.33    40
22)    45767    46658    0:00:29.763    STRICT_COMMERCIAL        1    0    0.00    -28.77    40
23)    46659    47553    0:00:29.863    STRICT_COMMERCIAL        1    0    0.00    -31.26    40
24)    47554    48470    0:00:30.597    STRICT_COMMERCIAL        22    0    0.00    -29.03    40
25)    *48471    49387    0:00:30.597    STRICT_COMMERCIAL        27    0    0.00    -28.11    40
26)    49388    49923    0:00:17.884    NOT_COMMERCIAL          35    0    0.00    -28.91    20
27)    49924    50827    0:00:30.163    STRICT_COMMERCIAL        5    0    0.00    -26.98    40
28)    50828    51236    0:00:13.646    NOT_COMMERCIAL          20    0    0.00    -29.45    20
29)    51237    51477    0:00:08.041    NOT_COMMERCIAL          5    0    0.00    -34.18    20
30)    51478    53311    0:01:01.194    NOT_COMMERCIAL          20    0    0.00    -27.52    20

In the snippet above, I have marked the frames with an asterisk where the actual commercial starts and stops. You can see that SA is determining that the commercial is running longer than it really is. I am a little bothered that it does not detect the channel logo (that's edge detection, right?) and wonder if that is part of the problem.

Anyhow, I am not sure how to tweak the settings to figure out just what is going wrong. I have tried changing a couple of things, but have not seen any improvement.

[andlju]

Absolutely. Have you tried the aggressive segmentation setting? It makes all the difference on a few channels in the UK. Just don't use it with cartoons.

Just curious.. What does "aggressive segmentation" mean? Or is that a trade secret..?

[Jere_Jones]

Just curious.. What does "aggressive segmentation" mean? Or is that a trade secret..?

Without it, a frame has to be dark and quiet to be considered a border frame. Aggressive Segmentation just takes it a little bit further so that frames that might not normally be considered as border frames are included.

Jere

[reboot]

I can't use agressive segmentation at all. It simply detects half the show as commercial, and half the commercials as show.
Without it, I have about 90% accurate detection, on Canadian stations, and slightly less on US stations.
Check the thread HERE for my settings, and the channels I've had most success on.

I simply need more tuners (2 PVR-500's aren't enough! ), and more time to work on settings for more channels.

@ Jere, not sure if you saw my latest post, but I changed drivers, and did get the audio thing working, although my logs show that the blackframe breaks are roughly as loud as the show's/commercial's are.
Show range: -22.20 to -26.10
Commercial range: -21.65 to -33.00

[Jere_Jones]

I can't use agressive segmentation at all. It simply detects half the show as commercial, and half the commercials as show.

Most people shouldn't use it. It is necessary, however, on some channels that NEVER have a blank frame.

I simply need more tuners (2 PVR-500's aren't enough! ), and more time to work on settings for more channels.

The next major addition will be a GUI for tuning the settings that let you see the results in real time.

@ Jere, not sure if you saw my latest post, but I changed drivers, and did get the audio thing working,

That's great. I am still going to investigate this. SA should be able to deal with those errors.

although my logs show that the blackframe breaks are roughly as loud as the show's/commercial's are.

The SPL column in the logs is the average for the entire block. The individual border frames themselves probably have a lower level.

Jere

[reboot]

The SPL column in the logs is the average for the entire block. The individual border frames themselves probably have a lower level.

Now it makes sense!
Show blocks are definitely lower (on average) than commercial blocks (usually).