Quickly find place/skip ads with binary chop

[mvallevand]

I don't see this being worth the trouble since whatever method you use is the accuracy of the first guess. Commercials here in Canada are almost always 30 seconds or 1 minute and breaks less than 8 minutes total so if you if you use splitting I see more keystrokes. 4 R R R R gets you to 8 you can go back 30 seconds at anytime. Because of the integer nature of each commercial you likey won't have those extra 7.5 seconds that the binomial approach gives in fewer keys.

Martin

[jcjefferies]

I tried Comskip several years ago and it was useless! It missed some adverts and detected adverts where there were none!

In the UK depending on the channel and time the adverts last around 4-5 minutes. In KODI I have skip set to +60 and -30 seconds. At the start of the advert 4 quick skips forward and if the adverts are still running a 5th skip. If the adverts have finished a skip backwards. So sometimes I see up to 30 seconds of adverts which I put up with for a simple solution!

Chris

[gEd]

I tried Comskip several years ago and it was useless! It missed some adverts and detected adverts where there were none!

In the UK depending on the channel and time the adverts last around 4-5 minutes. In KODI I have skip set to +60 and -30 seconds. At the start of the advert 4 quick skips forward and if the adverts are still running a 5th skip. If the adverts have finished a skip backwards. So sometimes I see up to 30 seconds of adverts which I put up with for a simple solution!

Chris

Yep, I have my npvr forward skip set to 60 seconds and 4 skips often takes me exactly to the end of the ad break.

I've tried it on and off since I started with GBPVR and never had any sucess with it in the UK

Yeh same here. Scanning of freeview HD content (H264) took ages (with the donator version of comskip) and it was never reliably accurate which makes it worse than not having it at all.
The problem in the UK is that there no consistent clear demarkation of when the ad starts (i.e black frames, silence or logos). In fact I seem to recall seeing in one show that the broadcaster had applied a comb effect to merge the last frame of the show and the start of the ad. I suspected that this was a deliberate attempt to twart ad skipping as there was a bit of a hu-ha about commercial PVR recorders being released with built in ad-skipping.

I think ads are much less of a problem for us in the UK anyway where we have an overall average of 7 minutes (max 12 mins) per hour, whereas in the US, a typical 30-minute block of time now includes 22 minutes of programming and eight minutes of advertisements [wikipedia]. Although I read that EU wants to get rid of this 12 min limit which I think is a crap idea. In fact, there you go, I just thought of a reason for Brexit, to preserve our great british limit of 12 mins of adverts per hour. That's got to be worth wiping out our economy and future economic prosperity for...!

[liteswap]

I tried Comskip several years ago and it was useless! It missed some adverts and detected adverts where there were none!

In the UK depending on the channel and time the adverts last around 4-5 minutes. In KODI I have skip set to +60 and -30 seconds. At the start of the advert 4 quick skips forward and if the adverts are still running a 5th skip. If the adverts have finished a skip backwards. So sometimes I see up to 30 seconds of adverts which I put up with for a simple solution!

Chris

+1

[pol098]

+1 for an optional binary search but I see one problem as described: Halving every skip after the first direction change could be quite bothersome. [Example] Nine keystrokes. The 5-second timeout would probably be used in this scenario. But when does one employ the timeout?

Answer to specific question: if you don't press an arrow for (say) 5", the binary chop resets itself. You also set a limit, I suggest also 5", which is the smallest chop - we stop halving there.

I posted this as someone who has used this algorithm, implemented as part of the Topfield Application Program (TAP) "QuickJump", independently developed by a member of the community of users of the Topfield TF5800 PVR (see http://www.toppy.org.uk/~simonc/ or search for <Quickjump Topfield>). I don't think source code has been published, but the binary chop is trivially easy to do - I've often used it. As a user I find it works extremely well. Pressing the same key a few times is just not a problem. But anyway it's such a simple process that it's quick and easy to experiment with settings and procedures.

I haven't found Comskip to be suitable for this job. I tried the third and last comskip.ini file on
http://www.kaashoek.com/comskip/viewtopi...f=2&t=1066 (but with detect_method=111 instead of 43), but didn't find it worked well - lots of false positives and false negatives. But even if it worked almost perfectly it would need tweaking, either at playback time (move or skip back and forth a bit to find right point), or when making a cleaned file. And the need to pre-process files with Comskip is another issue. Plus the binary chop isn't limited to ad-skipping - you can find a particular point in the file, if you know the contents fairly well, quickly.

If someone would implement this, it could be tweaked. I think there is too much reading of this thread in a theoretical way; you need to experiment with binary chops (in all sorts of situations) to appreciate how practical they are. (Try it -it's dead easy to write.) The QuickJump TAP implements a range of different fixed jumps (accessed by the red, green, blue, yellow, and white keys) on the remote controller, but |I always find myself using the arrows for binary chop on the PVR. Unfortunately the Topfield TF5800 is obsolescent, and as far as I know no other PVR has since been produced that allows the user community to write software for it.

As a trivial example of the power of the chop, I taught my young son to get someone to think of a number less than 1000 (1024 actually) and he would guess it in not more than 10 tries ([687] Try 500: is it correct, less, or greater? 750? 625? ... Impresses people. Requires a bit of mental arithmetic, so it was both fun and educational. And you can do any number less than 1,000,000 in 20 goes).

HTH

[pol098]

I don't see this being worth the trouble since whatever method you use is the accuracy of the first guess. Commercials here in Canada are almost always 30 seconds or 1 minute and breaks less than 8 minutes total so if you if you use splitting I see more keystrokes. 4 R R R R gets you to 8 you can go back 30 seconds at anytime. Because of the integer nature of each commercial you likey won't have those extra 7.5 seconds that the binomial approach gives in fewer keys.

I don't really understand the first sentence here. If you're using the chop for ad-skipping (it has other uses too) and commercials are reliably and exactly multiples of 30", then you could just press the skip key 8 times for a 4' break and land bang on the right spot. But where I am breaks aren't a multiple of 30", so successive jumps of 30" leave me beyond the end of the break, then I have to go back 30", then fast-forward carefully to find the right spot.

I do stress that this is something I've done in real life; the QuickJump method I use (see another very recent post of mine for details) gives me buttons for 5 different fixed skips as well as the chop; I always end up doing the chop because it's more convenient.

[NumberFive]

Shouldn't this be called "Successive approximation"?

https://commons.wikimedia.org/wiki/File:...aviour.png

[pol098]

Shouldn't this be called "Successive approximation"?

It's universally called "binary chop".
https://www.doc.ic.ac.uk/~kb/REASONING/J...up0809.pdf
http://www.doc.ic.ac.uk/~ar3/lectures/In...cture9.pdf

There seems little point in discussing what it should be called!

By the way, these references, from Imperial College, are useful for anyone planning to implement this, they say: "This is a very useful algorithm,
but one that is difficult to get right first time."

Best wishes

[NumberFive]

https://en.wikipedia.org/wiki/Binary_search_algorithm

In computer science, binary search, also known as half-interval search[1] or logarithmic search[2] and often called binary chop[3]is a search algorithm that finds the position of a target value within a sorted array.

Reference [3] is just Imperial College again, which I note you just added. Looking at some text books, they seem to agree it is a "binary search" or "half-interval search" which is much more descriptive.

My two cents, I use the skip buttons for +60 / -30 seconds and then the FF/RW buttons for +/- 10 seconds.

[jksmurf]

I tried Comskip several years ago and it was useless! It missed some adverts and detected adverts where there were none!
Chris

I've actually found it to be quite good, after some work, so I have persisted with it.

Occasionally I have had to send examples to Erik but he invariably comes back with suggested changes to the comskip.ini to make it work. There are a lot of methods available to pick up the ad breaks so some patience is required finding which works for you.

Sometimes it is actually excellent; the only bugbear is when the parameters on which the 'excellent' cuts work change, which is where it falls down. Then it's back to finding another method.

Like I said, my experience over the years has been pretty positive overall.

k.