questions relating optimization  

Share your tips and tricks here or get help with any Autopano Pro / Giga problem!
No bug reports (of any kind) in this forum!
no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

questions relating optimization

by klausesser » Thu May 01, 2014 11:28 am

Hi!

I´m wondering somebody really understands the way to optimize . .

let me start:
"Local approach: intelligently validates the local approach calculations. This changes in the assembly if the calculations seem wrong (false links)"
What does it mean in detail: "This changes in the assembly if the calculations seem wrong (false links)"?

"Grid position":
"Use grid position: only available if you have imported pictures taken with the help of a pan head via a Clauss or Papywizard import plug-in (this option re-uses the coordinates provided to pre-position the images)."

So this setting is to be used always with using a motorized heads providing xml?

"Clean"
When i want to have a RMS of max. 1 to 1.3: what do i need to set? I have no idea!

"Optimization scopes"
I read the "optimizer engine documentation" - what´s it supposed to tell me?

What do i need to do when i want RMS lower than 1.3?

I use a very precise head - max offset is 0,036°/10000steps. Importing xml and images with
having set "skip optimization" gives me a very good result already.
But of course it needs to be optimized nevertheless.

I new how to use it successfully in 3.0.8.
I do not know how to use it at all in 3.5.1

So: in cases the automated optimization isn´t good enough - which way to go (without endlessly try and error)?

"Image":
"the related setting would be applied on individual images" what does "individual images" mean in detail? Which images do you mean?

1) my Nikon D800 writes EXIFs for the body - always.
2) it also writes EXIFs relating to the focal-length and aperture for my AI/AIs lenses which values i type into the camera.
So: which EXIFs are taken for "Grouping Criteria"?

Is it preferable to always use "Automatic" for "optimization scoppes"?
In which cases might it be preferable to use different settings?

Fact is: i have issues using the vertical lines tool effectively as long as the stitch is not correctly optimized.

So: how can i optimize correctly?

Klaus

User avatar
Destiny
Moderator
 
Posts: 4992
Joined: Mon Apr 26, 2010 11:55 pm
Location: Australia

Re: questions relating optimization

by Destiny » Thu May 01, 2014 11:39 am

Some Optimisation info..

http://www.autopano.net/wiki-en/action/ ... timization

Use the sliders to control the RMS values...

Destiny..

User avatar
AntoineM
Kolor Team
 
Posts: 69
Joined: Sun Apr 21, 2013 6:43 pm

Re: questions relating optimization

by AntoineM » Thu May 01, 2014 12:13 pm

Hello klausesser,

Your questioning about optimization scopes is very reasonable as I just noticed we have some inconsistencies about this specific point in the documentation page you quote ("Autopano - Preferences... - Optimization").
While we are working on clarifying this page, I suggest you to study the explanation given in this other one: http://www.autopano.net/wiki-en/action/ ... zer_engine

Getting an RMS lower than 1.3 is very easy, you just have to clean more points.
Getting a good stitch is different: you have to clean only bad points and share unknowns that must be shared.
This is explained in this documentation page: http://www.autopano.net/wiki-en/action/ ... _RMS_value

Best regards,
Antoine

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Thu May 01, 2014 12:28 pm

AntoineM wrote:Getting an RMS lower than 1.3 is very easy, you just have to clean more points.


Hi Antoine!

That´s really a brilliant idea - never would have guessed . . =D :cool:
How? Point by point in the CP editor window? What about having some hundred images?

AntoineM wrote:Getting a good stitch is different: you have to clean only bad points and share unknowns that must be shared.
This is explained in this documentation page: http://www.autopano.net/wiki-en/action/ ... _RMS_value


No. It´s not "explained" there in a way an average user can understand it. Guess i´m a bit more than an "average" user - and i
do not understand it.

AntoineM wrote:"you have to clean only bad points and share unknowns that must be shared"


Aha?

Klaus

no avatar
mediavets
Moderator
 
Posts: 14161
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.

Re: questions relating optimization

by mediavets » Thu May 01, 2014 12:41 pm

Kaus,

I think AntoineM's answer was perhaps deliberately somewhat teasing/provocative when he says 'Getting an RMS lower than 1.3 is very easy, you just have to clean more points'.

I think he's saying that there's a difference between reducing RMS to below 1.3 and achieving a good stitch.

As I understand it optimising in APG is not the same as optimising in PTGui, in the latter reducing the 'magic number' - forgotten what their statistic of goodness of fit is called - is always deemed to be better.

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Thu May 01, 2014 2:16 pm

mediavets wrote:Kaus,
I think AntoineM's answer was perhaps deliberately somewhat teasing/provocative when he says 'Getting an RMS lower than 1.3 is very easy, you just have to clean more points'.


Fact is: having a good RMS you have a good stitch. My praxis - PRAXIS - shows that i NEED a RMS of max. 1.3 for having an error-free stitch with 700mpx and higher when zooming in to 100%. What´s your average size of equis, Andrew and Antoine?

mediavets wrote:I think he's saying that there's a difference between reducing RMS to below 1.3 and achieving a good stitch.


I know what i do, Andrew - and i know what i need ;) :cool: Shooting and stitching a pano using a fisheye is easy (well: more or less, depending on the stitcher =D )
But shooting hires panos with 42 or 214 images is quite different edit-wise. Here i realized - you will not be able to prove the opposite - that
i need to have at least RMS of 1.3 for having NO errors when zooming to 100%.
RMS 2 for example needs me to critically investigate the *whole* equi in Photoshop for hours for finding errors and retouching them.

mediavets wrote:As I understand it optimising in APG is not the same as optimising in PTGui, in the latter reducing the 'magic number' - forgotten what their statistic of goodness of fit is called - is always deemed to be better.


In PTGui i don´t have these issues. PTGui provides magnifying tools for editing on images for seeing what happens. It lets me pixel-precise edit CPs manually in realtime @100%.
But that´s rarely needed. Good to have anyway. IF there ever are HEAVY issues with CPs i can use the mask-tool for masking bad areas in overlaps - this way
ANY stitch can be done perfectly.
I needed to do it only twice - when i made mistakes by not tightening the screw for the camera-rail on the head when it took the Zenith shot: the rail slipped two centimeters when the head shot the upper row and then did the Zenith shot:
35mm, 42 shots, 20% overlap indoors: http://www.360impressions.de/Klap - multires on tablets also.

I usually do not at all count on numbers alone. But there definitely is a coherence between the number of RMS and a good stitch.
What might RMS values tell us otherwise??

Again: a pano with 6 or 7 fisheye shots using a crop camera can work well with RMS of 2.5 or even 3 - you simply do not see minor erors even at 100%. But use a D800 with a 16mm fisheye providing 180MPx . . . and the whole thing looks very diffenent @100% - believe me . . :cool: :cool:

Again #2 as said already: a D800 with a 35mm lens provides about 1,2GPx. Looking on the stitch @ 100% will teach you to think about it totally different . . .

Maybe even in that case RMS of 2 or more MIGHT work without errors - i don´t know - but you CANNOT BE SURE it does.
Producing commercially/professionally i NEED to be sure to work with NO visible errors as far as it´s possible!

Ain´t not too hard to accept, is it . . :cool:

Klaus

User avatar
Zoobius
Kolor Team
 
Posts: 27
Joined: Wed May 19, 2010 1:23 pm
Location: Kolor, France

Re: questions relating optimization

by Zoobius » Fri May 02, 2014 1:28 pm

So.. here is the tricky way to get the RMS value you want:

• activate the control points tool
• select the Control points window and focus on the Selection lists: http://www.autopano.net/wiki-en/action/ ... tion_lists
• then open the list of Control points
• sort the points by RMS to get the worst RMS values on the top of the list
• select all the values over 5 (maybe 4 or less... for the challenge but I don't guarantee anything with less than 5) and delete them
• re-optimize your panorama

Now you should get your perfect RMS, if not, decrease the values to remove and jump to 3 or less but that would be really challenging!

Finally I advise you to check the final image at 100% anyway, to be sure your aim is reached ;)
Guillaume - Kolor Technical Support

Quick links: User Account, Documentation, Youtube Channel

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 2:11 pm

Zoobius wrote:So.. here is the tricky way to get the RMS value you want:

• activate the control points tool
• select the Control points window and focus on the Selection lists: http://www.autopano.net/wiki-en/action/ ... tion_lists
• then open the list of Control points
• sort the points by RMS to get the worst RMS values on the top of the list
• select all the values over 5 (maybe 4 or less... for the challenge but I don't guarantee anything with less than 5) and delete them
• re-optimize your panorama

Now you should get your perfect RMS, if not, decrease the values to remove and jump to 3 or less but that would be really challenging!


Hey Guillaume!

That´s the way i work anyway when i have some few images to stitch :cool:
But can you imagine what that means having 300 or 700 or more images?? =D

No - this can´t be the way to go obviously.

I guess you misunderstood what i mean: I only want to know what i need set in the CP-editor of 3.5.1 like i did in 3.0.8.
In the "old" editor i could klick on "delete wrong CPs". This option doesnt´t exist in 3.5.1 - or it´s in another place and/or
is named different.

I JUST CAN´T FIND this option!!

I need to be able to set: "delete all CPs with more than RMS 2" - that woulkd be ok. After having done so i want to set: "delete all CPs with more than
RMS 1.5"

Again: i guess this option exists already in 3.5.1 . . . where?

I read about "scopes" and whatever . .

Of course i tried the "filter" option . . . but it does not do what i´m used to.

So i´m really helpless about how to use the surely ;) countless brilliant new features ;) in 3.5.1 - if i only would know where they are and how they work in detail and how to use them constructively.

The "documentation" is in a way like describing how to fly an airplane by telling that it has a motor . . . wow. =D :cool:

Zoobius wrote:Finally I advise you to check the final image at 100% anyway, to be sure your aim is reached ;)


REALLY?? Well - now that you´re telling me . . . . =D :cool:

Klaus

User avatar
Zoobius
Kolor Team
 
Posts: 27
Joined: Wed May 19, 2010 1:23 pm
Location: Kolor, France

Re: questions relating optimization

by Zoobius » Fri May 02, 2014 2:49 pm

klausesser wrote: But can you imagine what that means having 300 or 700 or more images?? =D


Does it mean you do it for each image separately? I hope not. Btw I forgot to mention to prior select all the images on the main top list to display all the links and control points on the list of Control points. And indeed you would have to scroll away to get all the points you want to remove with 300 images I agree… but this is not the only way to do it!


klausesser wrote: In the "old" editor i could klick on "delete wrong CPs". This option doesnt´t exist in 3.5.1 - or it´s in another place and/or is named different.


Yes we changed the interface... but don't worry! You can still filter automatically the points (yes the previous trick I proposed was an other way to do it as lists are not known enough, this is sometimes good to show such interesting details up, isn't it?)

So, in the Filter tab there is 2 options, Filter links and Filter points. Set the value you want on the Filter points option then hit the icon to clear points over the value you set, that's all!
Guillaume - Kolor Technical Support

Quick links: User Account, Documentation, Youtube Channel

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 3:02 pm

Zoobius wrote:So, in the Filter tab there is 2 options, Filter links and Filter points. Set the value you want on the Filter points option then hit the icon to clear points over the value you set, that's all!


Yes, I know. I figured it out of course. But: it doesnt´work reliably as it seems to me.
Having used the filter the CP values didn´t change.

Guillaume: please don´t think i´m an idiot . . =D :cool: I usually work with 700mpx to 6gpx panos twice a week for years now.

So far i will not use 3.5.1 for working on jobs. Need to dig completely new into APG first as it seems.

Afaik other commercial producers of panos metioned a broken consistency in the use of APG also - right, Hans?

Klaus

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 3:29 pm

Zoobius wrote:So, in the Filter tab there is 2 options, Filter links and Filter points. Set the value you want on the Filter points option then hit the icon to clear points over the value you set, that's all!


Could you please show me "the icon to clear points"? I find icons for "full optimize" and "fast optimize". No icon for deliberately "delete bad points".

klaus

PS: which "filtering model" would you suggest in which case?

PS: sorry - i was looking in the "points" tab for the "delete points" icon!

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 3:36 pm

Ok - i used the "filter" set to "2" and hit "clear points".

Result: ALL points are gone - there were links/points with 4,6 and 2,8 . .

Klaus
Last edited by klausesser on Fri May 02, 2014 3:38 pm, edited 1 time in total.

User avatar
Zoobius
Kolor Team
 
Posts: 27
Joined: Wed May 19, 2010 1:23 pm
Location: Kolor, France

Re: questions relating optimization

by Zoobius » Fri May 02, 2014 3:37 pm

I don't think you are an idiot! I'm there to help you as well as I could help any other forum member. If you are aware about that feature, that could be not the case for everyone. And because you said you didn't find it, I simply pointed it for you ;)

Let's continue with the RMS issue. Did you check the Change log of the 3.5 release? Here is what we said about it:

"RMS values. This value is the standard way to measure if a panorama is well stitched or not. Or you'll probably see that values of RMS in version 3.5 haven't really changed whereas the quality has increased a lot ( if you compare visually both 100% rendering from 3.0 and 3.5 ). This is normal as we changed the way RMS is measured. So RMS=4.0 in version 3.5 is better than in version 3.0. Nevertheless, there are quite comparable, but if you found higher RMS panoramas now, don't assume the panorama is worth. Most of the time it is better."

And the icon is there:
Attachments
filter-points-icon.jpg
filter-points-icon.jpg (10.67 KiB) Viewed 327 times
Guillaume - Kolor Technical Support

Quick links: User Account, Documentation, Youtube Channel

User avatar
Zoobius
Kolor Team
 
Posts: 27
Joined: Wed May 19, 2010 1:23 pm
Location: Kolor, France

Re: questions relating optimization

by Zoobius » Fri May 02, 2014 3:57 pm

klausesser wrote: which "filtering model" would you suggest in which case?


The automatic mode is the better, as it smartly uses some parts of each model available depending on the calculations, what is impossible to do by setting a specific model.

klausesser wrote: Ok - i used the "filter" set to "2" and hit "clear points".

Result: ALL points are gone - there were links/points with 4,6 and 2,8 . .


Setting filter values to 2.00 if the lower value you have is 2.8 would indeed remove all the points!
Guillaume - Kolor Technical Support

Quick links: User Account, Documentation, Youtube Channel

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 3:59 pm

Zoobius wrote:So RMS=4.0 in version 3.5 is better than in version 3.0. Nevertheless, there are quite comparable, but if you found higher RMS panoramas now, don't assume the panorama is worth. Most of the time it is better."



Well - but there´s no reliability as i t seems: "most of the time it is better" . . . that means "maybe, maybe not".
That´s not a reliable way to do work! It´s surely ok for playing around.

Again: for producing panos i need to know "what happens when - and why does it happen". And i need to know
how to achive excellent results in an appropriate time without needing to try and error around all night long for finding the optimal way to go.

Again: i have an RMS of 4,3 to 2.8 in 6 links. Now i set the "filter points" to "2" or "3" and hit "delete points"- and ALL links are gone. Why´s that?

Shouldn´t only points above 2 or 3 get deleted? In trhe editor window i see offsets. Hitting "preview" its better -but there still are offsets.

And this happens with only 6 shots - imagine it´s 600 . . .

Again once more: i need to have a reliable way to edit panos with very much images in a way i DO NOT need to investigate the final rendering
for offsets more than in the way i look go for errors anyway. It takes HOURS investiganting a 700mpx or 6gpx pano looking for stitch errors.

Si ie need to have a reliable way to estimate the amount of errors i must expect - for knowing whether i need to edit more . . BEFORE i render the result and THEN
realize "oops - there still are errors . . need to do it al again because the software didn´t show me or informed me about the precision of my stitch".

Really: that´s not what i like at all!

Klaus

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 4:00 pm

Zoobius wrote:
klausesser wrote: which "filtering model" would you suggest in which case?


The automatic mode is the better, as it smartly uses some parts of each model available depending on the calculations, what is impossible to do by setting a specific model.

klausesser wrote: Ok - i used the "filter" set to "2" and hit "clear points".

Result: ALL points are gone - there were links/points with 4,6 and 2,8 . .


Setting filter values to 2.00 if the lower value you have is 2.8 would indeed remove all the points!


Why? I also set it to "3" - same effect. What about the 4,6 value? A seting of "3" shouldn´t delete ALL points, should it?

Klaus

PS - i mean: what good are values anyway when we can´t rust on them??

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 4:23 pm

Now:

i started the app again and set "filter points" to 3 again. Hitting "clear points" now show RMS values of about 2,9 to 1,7 as blue icons and 1,8 to 1,5 as green icons when i optimize the whole stitch.

Half an hour ago i did the same thing and also using "3" - and ALL points were gone, as already said.

Before i did use "3" i used "2" - and went back one step for doing it again using the "3" value. Could this cause an issue: using "undo", going back and use a different value then?

I have no explanation why the same acting results in different results suddenly - after just closing and restarting the app.

Klaus

User avatar
AntoineM
Kolor Team
 
Posts: 69
Joined: Sun Apr 21, 2013 6:43 pm

Re: questions relating optimization

by AntoineM » Fri May 02, 2014 4:32 pm

Hello Klaus,

If you did:
* clean with limit=2
* optimize
you had to go two steps back before trying a new clean limit of 3: one to undo the optimization and one to undo the clean.

Best regards,
Antoine

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 4:34 pm

even more funny: i closed the app and started it again. Used the same xml/image import as before. Now the importer´s "common" tab tells me:
"lens type difference between EXIF and XML"

But in the previous run some 15 minutes ago this message did NOT show up!

Nevertheless: the "image tab" shows me the correct focal and lens model - which in the previous "common" tab was stated
faulty . .

Klaus

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 4:35 pm

AntoineM wrote:Hello Klaus,

If you did:
* clean with limit=2
* optimize
you had to go two steps back before trying a new clean limit of 3: one to undo the optimization and one to undo the clean.

Best regards,
Antoine



I didn´t use "optimize" but i used "clean points"! But you´re right - forgot to mention i always go back two steps when using a certain command in between.

best, Klaus

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 4:54 pm

Ok: the issues clearly sat in front of the screen . . . =D :cool: :rolleyes:

problems solved! Guess i now learned how to use it all.

When you specialize on a certain way to handle things it sometimes is very irritating when a way to handle
changes - and you don´t have much time to evaluate all the changes in deep.

Thank you all for caring!

best, Klaus

no avatar
klausesser
Member
 
Posts: 7858
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany

Re: questions relating optimization

by klausesser » Fri May 02, 2014 5:20 pm

Sorry: nothing is good . .

After half an hour i started APG 3.5.1 again and used the same import which i poated for download here.

The result after importing and first step of optimizing:
Bildschirmfoto 2014-05-02 um 18.10.42.png


But don´t be afraid: at the moment i don´t have time anymore to discuss it :cool:

best, Klaus


Return to Using Autopano Pro & Autopano Giga

Who is online

Users browsing this forum: No registered users and 2 guests