ShowAnalyzer Beta

[Jere_Jones]

Regarding the virus warnings, Avast has responded saying that they are false warnings and the virus definitions will be updated.

The post is here.

Thanks,
Jere

[reboot]

I was just going to post. I (finally) got the email back from Avast, and they'll have updated def's later today.

[Spartan]

Just wanted to report that I am having mixed results with SA v0.5.0. I have started running the "Into the West" series from TNT (USA) through it and episode 6 & 1 came out flawless, episode 2 had 2 minor false positives (easily removed with VideoRedo) and episode 3 had all sorts of false positives. Regarding #3, I made to make all all of the cuts by hand as SA just couldn't seem to figure that one out. It seems that SA liked to call many of the dark scenes commercials.

I have not yet examined the output from episodes 4 & 5.

All of this was done with the default settings.

Let me know if you are interested in logfiles or anything -- of if there are some other settings I should try.

-S

[reboot]

You need to look at the end of the log, in the Block List.
Open the mpg in virtualdubmod.
Find the start of the NOT detected commercial.
Note the frame number.
Look for that frame number in the log.
It will say something after it, like, "NON_STRICT_COMMERCIAL".
It will also give a number at the end, the "Score".
Very low numbers are show, larger numbers are commercials.
You can then tweak a few settings, depending on what SA says about that particular block.
I had to do a bit of tuning as well, but nowhere near as much as the comskip tuning I went through!

[ydekmekji]

Where did you find the log files? In the folder, there are only comskip logs--which makes me wonder... I edited comskip out of my postproecessing...

[smeghead]

Hi All

I'm having problems understanding how this works. If the score is more than 1 then its a commercial right? that isn't what I have found as shown below.

I have used a 40 minute program to illustrate. Having looked at the program in a video editor the commercial breaks are as follows:
0 - 1337
22127 - 29867
52532 - 61947 (The following show then starts at 61947)


The SA v0.5.1. result txt file contained the following:
0 319
22950 23962
48805 49054
56294 58066
59336 60865


The Block list at the end of the log file was:

Block List
-----------------------------------------------------------------------------------------------------
# Start End Length Eval of Length Blanks EP SC AvgSpl Score
-----------------------------------------------------------------------------------------------------
Commercial
0) 0 314 0:00:12.599 NOT_COMMERCIAL 0 1 0.00 -35.48 0.7
Show
1) 315 576 0:00:10.480 NON_STRICT_COMMERCIAL 11 1 0.00 -47.09 1.5
2) 577 838 0:00:10.480 NON_STRICT_COMMERCIAL 12 1 0.00 -30.59 1.5
3) 839 7997 0:04:46.360 TOO_LONG 1 1 0.00 -44.05 0.1
4) 7998 22179 0:09:27.279 TOO_LONG 0 1 1 0.00 -42.52 0.1
5) 22180 22944 0:00:30.600 NOT_COMMERCIAL 1 1 0.00 -30.87 0.7
Commercial
6) 22945 23956 0:00:40.479 NON_STRICT_COMMERCIAL 11 0 0.00 -37.22 30
Show
7) 23957 24718 0:00:30.480 STRICT_COMMERCIAL 12 1 0.00 -36.02 2
8) 24719 25480 0:00:30.480 STRICT_COMMERCIAL 12 1 0.00 -38.35 2
9) 25481 26242 0:00:30.480 STRICT_COMMERCIAL 11 1 0.00 -37.79 2
10) 26243 26754 0:00:20.480 STRICT_COMMERCIAL 11 1 0.00 -32.75 2
11) 26755 27766 0:00:40.479 NON_STRICT_COMMERCIAL 12 1 0.00 -33.40 1.5
12) 27767 28028 0:00:10.480 NON_STRICT_COMMERCIAL 12 1 0.00 -38.96 1.5
13) 28029 28290 0:00:10.480 NON_STRICT_COMMERCIAL 12 1 0.00 -44.70 1.5
14) 28291 29052 0:00:30.480 STRICT_COMMERCIAL 11 1 0.00 -35.44 2
15) 29053 29552 0:00:20.000 STRICT_COMMERCIAL 12 1 0.00 -31.21 2
16) 29553 29878 0:00:13.039 NOT_COMMERCIAL 1 1 0.00 -35.61 0.7
17) 29879 46519 0:11:05.639 TOO_LONG 1 1 0.00 -43.14 0.1
18) 46520 48804 0:01:31.400 TOO_LONG 1 1 0.00 -46.31 0.1
Commercial
19) 48805 49054 0:00:10.000 NON_STRICT_COMMERCIAL 1 0 0.00 -39.84 30
Show
20) 49055 54763 0:03:48.360 TOO_LONG 1 1 0.00 -38.35 0.1
21) 54764 55157 0:00:15.759 NOT_COMMERCIAL 12 1 0.00 -34.20 0.7
22) 55158 55525 0:00:14.720 STRICT_COMMERCIAL 1 1 0.00 -33.58 2
23) 55526 56287 0:00:30.480 STRICT_COMMERCIAL 12 1 0.00 -34.61 2
Commercial
24) 56288 57049 0:00:30.480 STRICT_COMMERCIAL 12 0 0.00 -37.56 40
25) 57050 57912 0:00:34.520 NON_STRICT_COMMERCIAL 12 0 0.00 -36.85 30
26) 57913 58006 0:00:03.759 TOO_SHORT 1 0 0.00 -34.24 20
27) 58007 58061 0:00:02.200 TOO_SHORT 1 0 0.00 -49.53 20
Show
28) 58062 58700 0:00:25.559 STRICT_COMBINED 11 1 0.00 -35.31 2
29) 58701 58823 0:00:04.919 STRICT_COMBINED 1 1 0.00 -36.75 2
30) 58824 59335 0:00:20.480 STRICT_COMMERCIAL 11 1 0.00 -35.71 2
Commercial
31) 59336 59585 0:00:10.000 NON_STRICT_COMMERCIAL 1 0 0.00 -38.10 30
32) 59586 60097 0:00:20.480 STRICT_COMMERCIAL 12 1 0.00 -40.42 2
33) 60098 60859 0:00:30.480 STRICT_COMMERCIAL 12 0 0.00 -34.14 40
Show
34) 60860 61753 0:00:35.759 NOT_COMMERCIAL 12 1 0.00 -35.28 0.7
35) 61754 63699 0:01:17.840 TOO_LONG 1 1 0.00 -34.88 0.1
36) 63700 63742 0:00:01.720 TOO_SHORT 1 1 0.00 -35.35 1
37) 63743 64623 0:00:35.240 NON_STRICT_COMMERCIAL 1 1 0.00 -34.28 1.5
Commercial
38) 64624 67271 0:01:45.920 TOO_LONG 1 1 0.00 -34.60 0.1
-----------------------------------------------------------------------------------------------------

So on to my questions:
1. Is a score of more than 1 a commercial
2. Why hasn't more blocks been identified as commercials in the above. Such as blocks 7 to 15?
3. What do the columns mean and how is the score actually calculated from them?
4. Will the built in decoder give better results than the free version

Thanks in advance for any help

[elite]

2) Is the total length longer than longest possible commercial segment in SA config?

Not sure about the others...

[Jere_Jones]

Where did you find the log files? In the folder, there are only comskip logs--which makes me wonder... I edited comskip out of my postproecessing...

They should be in the same folder as the analyzed mpeg. But you'll need to turn them on in SAConfig. Just put a check beside "Save log to disk".

Jere

[Jere_Jones]

So on to my questions:
1. Is a score of more than 1 a commercial

No. In general, if the previous block was a commercial, a score below 15 will be marked as a show. If the previous block was a show, a score above 10 will be marked as a commercial.

2. Why hasn't more blocks been identified as commercials in the above. Such as blocks 7 to 15?

They seem to all have logos which increased their score accordingly.

3. What do the columns mean and how is the score actually calculated from them?

# - Unique sorted identifier
Start - The start frame
End - The end frame
Length - The length
Eval of length - What the length was judged as
Blanks - How many border frames start the block
EP - Which edge profile that block matches (-1 means unknown, 0 means none, 1 or greater indicates a specific edge profile)
SC - Average scene change (inactive)
AvgSpl - The average audio level of the block
Score - The sum of all known characteristics of the block
The score calculation is straight forward. Take a look at the score modifiers in SAConfig. If a block matches the category, the score is multiplied by the modifier.

4. Will the built in decoder give better results than the free version

On normal SD MPEG2 recordings the free decoders and the built in decoders should provide the same results (in theory). On HD files with AC-3 audio, the built in decoders will give superior results.

Jere

[smeghead]

2) Is the total length longer than longest possible commercial segment in SA config?

Not sure about the others...

I've tried lots of settings, those I posted above gave the best results. The default settings didn't get any commercials. Anyway, I took your advice from an earlier post and my settings are almost the same as yours, they are:

Longest Possible Commercial Segment: 400
Longest Possible Individual Commercial: 60
Shortest Possible Commercial Segment: 10
Shortest possible Show Segment: 420

Length Too Long: 0.1
Length Not a Commercial: 0.7
Length Too Short: 1
Length Nonstrict Combined: 1.2
Length Nonstrict: 1.5
Length Strict Combined: 2
Length Strict: 2
Edge Profile Positve Match: 0.1
Edge Profile Negative Match: 2
Edge Profile Unknown: 1
Aggressive Segmentation: Yes

Also, why is the average audio level negative?
And does SA use a dictionary like comskip?

Ooops - it's a 45 minute program, forgot about the pre and post padding. Not that it makes a differnece but I just wanted to be accurate