about ping-pong prediction.
 

hello, I have read all the thread written from incredible:

I have created a avi of 30m in order making of the tests.
I have used:
slicer (2,15,0,2,1) for ping
slicer (2,15,30,2,1) for pong

with ping I have 4274 kb
with pong I have 4358 kb

medium value is of 4316 kb

previewed filesize 4311*100/2 = 215,800 kb

final resulting has been of 196.172 kb (less of the 9,09 %).

I would want to know if all this is right and that precision has the method ping-pong.

thanks

Re: about ping-pong prediction.
 

:?:

Maybe I cant follow but ...


In your case 4316kb would end in 215800 kb, ok ... how do you come to 4311?

Maybe you should try:

Ping : Slicer(2,15,05,2,0)
Pong: Slicer(2,15,30,2,0)

As by this you can avoid predictionning the first 5 Mins where almost every Movie in its "intro" got very much dark.

Maybe even Ping-Pong did not catch a very dynamic scene in your movie, ... so you can assume what would happen if you would have predicted the "regular" way 8)

Your difference in 800 MB file, it woul be about 60MB. It is to much difference.
Do you run a safe ping test?
Could you run ping pong prediction with slicer function Prodater64Mod (look at end of the pingpong thread) and compare with your results?

reply
 

@prodator64
I would want to understand the incredible method's and then using your modified ruotine, otherwise I risk not to understand nothing.

@incredible
you can explain in way simple and schematic all the passages to me in order to apply your method?

thanks

Heres everything explained incl. the function, pics and all you need:
http://www.kvcd.net/forum/viewtopic.php?t=7626&start=1
:)

reply
 

hello incredible, I would want to know some things:

1. If I use a percentage of 5% instead of 2% for the prediction, I will have a best prediction?

2. the recommendation in order to find the offset one to parity of gop and gl depends by sa, that is if I use 5% offset will be always 12, right?

3. You can tell me, large way, that error percentage you have, as an example

I have: sampler(15) ---> 6025 kb
previewed Kb 602500
real Kb 551399 kb

error with normal prediction (+8,48%)
error with ping pong (2%): (+4,11%)
error with ping pong (5%): (+2,43%)

are in the average?

by

1) if using 5% at ping-pong you in real do predict 10%!! of the movie as you do predict using 2 offsets ... means for example 05 and 35.
A prediction using ping pong of 2% is TOTALLY enough ;-)

2) Waht do you mean?? I didnt understand.
The internal calculator does its calcualtion in a way that the samples of the 2nd offset (pong) always will be almost in the middle between the ping samples (see my green graphic in the link of the "long-short vs. ping pong" thread)

3) Again I didn't understand.
I do always use 2% and sometimes a gl of 1 and sometimes a gl of 2 , so try a goplenght multiplicator (gl) of just 1 (offset decreases automatically) and tell me what results.
As told I got everytime about 98% acuracy.

then, my objective is this:

to make the prevision for 5 films together:

using the normal method (that is sampler(lenght=15)) I get a total prevision regarding the real value of the +4,54%

using your method with
slicer(2,15,00,1,0) for the ping and
slicer(2,15,15,1,0) for the pong
I get the prevision of the +5.91% regarding the real value.

I use like encoder cce.

where's the mistake?

Means?

Do you say "Sampler(15) does it better or what?
Because as seen above slicer(2,15,00,1,0) does also sample eoth a sample length of 15!

Maybe you should try using slicer(2,15,00,2,0)
And therefore the higher Offset in Pong i.E. slicer(2,15,35,2,0)