|
From: skunkwerk <sku...@gm...> - 2008-07-22 16:31:29
|
Hey Sebastian,
this one worked for me: www.akbars.net/10.zip
i've been testing with collada models from
sketchup.google.com/3dwarehousebut that's because my version of the
collada importer works ok with the
textures - i don't think the one in subversion has been fixed yet...
try increasing the size of the texture atlas until i can get the resampled
tiles working. i'm going to test out the results on a simple box today
paolo, the rect_packer.cpp has a bunch of asserts in it... i'd like to
change them to return false, so that the atlas will fail, but not crash
whenever some of the images are null, etc. the other way would be to try to
do those checks in the filter itself. which would you recommend?
imran
On Tue, Jul 22, 2008 at 3:25 AM, Sebastian Herholz <
seb...@go...> wrote:
> Hi Imran,
>
> I tried the Build Texture Atlas filter and I always get the message Failure
> of Filter: "Build texture atlas".
> Can you host somewhere the objects that worked for you?
>
> bye
>
> Sebastian
>
> 2008/7/22 skunkwerk <sku...@gm...>:
>
> team,
>> i've fixed 2 rather insidious logic bugs in my texture atlas and
>> committed the changes:
>>
>>
>> - uv coordinates need to be converted to their corresponding positive
>> values BEFORE the triangle spans are calculated
>> - i wasn't storing the maximum span, but rather the span of the last
>> triangle to be iterated... dumb mistake
>>
>> i've tested on about 10 models, it hasn't crashed (at least with the
>> default 'copied tiles'), and the UV coordinates are now always between 0 &
>> 1. in 2 of the cases it failed because the textures couldn't fit in the
>> default size due to tiling.
>>
>> i'll try some simple models (a box with 6 textures) to make sure the
>> mapping is correct...
>>
>> 1) can someone confirm that a uv coordinate of -1.6 is equivalent to a
>> positive uv coordinate of 1.4?
>>
>> thanks,
>> imran
>>
>>
>> On Sun, Jul 20, 2008 at 4:11 PM, skunkwerk <sku...@gm...> wrote:
>>
>>> thanks paolo... sorry, that was a typo. the exact same code works in the
>>> colladaformat.h file, but not in my filtertexture.cpp file - any idea why?
>>>
>>>
>>> On Sat, Jul 19, 2008 at 9:57 PM, Paolo Cignoni <
>>> pao...@is...> wrote:
>>>
>>>> In this mail you wrote
>>>>
>>>> CMesh<zero> instead of CMesh<Capital O>
>>>>
>>>> check that you did not the same in the code...
>>>>
>>>> p.
>>>>
>>>>
>>>>
>>>> skunkwerk wrote:
>>>>
>>>> yeah, i read the wiki and looked at my own source code.
>>>>
>>>> this used to work as of a few months ago:
>>>> vcg::tri::Allocator<CMesh0>::CompactVertexVector(m);
>>>> but isn't anymore...
>>>>
>>>> the compiler says the candidate function is the same thing...
>>>>
>>>> 2008/7/17 Paolo Cignoni <pao...@is...> <pao...@is...>:
>>>>
>>>>
>>>>
>>>> skunkwerk wrote:
>>>>
>>>> i've updated and committed the texture atlas filter
>>>>
>>>> 1) i've removed the extraneous code so it doesn't bother trying to create
>>>> multiple atlases now
>>>> 2) i make any negative UV coords positive before I try to offset them, as
>>>> it's easier to figure out that way. i've also fixed some code in
>>>> adjustUVCoords so it's a bit more straightforward now
>>>> 3) i couldn't find the "updateTexture" slot you mentioned, Paolo. did you
>>>> mean the updateGL() slot in mainwindow?
>>>>
>>>>
>>>> it is a slot of the GlArea not of the main window. So it could be an issue
>>>> since from within a filter you should not know a lot of the glarea....
>>>> I will add a solution to this.
>>>>
>>>> 4) i'm still not sure what the correct syntax is for compactfacevector,
>>>> etc. so it's commented out right now
>>>>
>>>> It is explained in the vcg wiki
>>>> http://vcg.sourceforge.net/index.php/Allocation
>>>>
>>>>
>>>> Cheers
>>>> p.
>>>>
>>>>
>>>>
>>>>
>>>> it compiles... but i'll test it tomorrow morning
>>>>
>>>> On Thu, May 22, 2008 at 3:42 PM, skunkwerk <sku...@gm...> <sku...@gm...> wrote:
>>>>
>>>>
>>>>
>>>> thanks Paolo,
>>>> i've commited a new version:
>>>>
>>>> 1) the reason for multiple texture atlases was that some platforms cannot
>>>> support very large textures - but i've changed the default texture size to
>>>> 4096x4096
>>>> 2) i've split up the code into 5 separate functions
>>>> 3) i'm using stl's min/max now
>>>> 4) i tried calling the compactface function but was getting an error and
>>>> have commented it out for now - what's the correct way to do this?
>>>> 5) i'm aware of the floor and ceiling functions, but didn't see any place
>>>> in the code i could have used it
>>>>
>>>> imran
>>>>
>>>> On Wed, May 21, 2008 at 11:03 PM, Paolo Cignoni <pao...@is...> wrote:
>>>>
>>>>
>>>>
>>>> skunkwerk wrote:
>>>>
>>>> i've made further changes to the texture atlas plugin - it should now
>>>> work with negative uv coords. it is crashing, however on line 197:
>>>>
>>>> it causes a segmentation fault, that i've backtraced to qpainter. This
>>>> does not crash when I do not try to scale the qimage before converting it to
>>>> a qpixmap. any suggestions? I've tried calling the qpainter destructor
>>>> before resizing, but that doesn't help.
>>>>
>>>> http://meshlab.svn.sourceforge.net/viewvc/meshlab/trunk/meshlab/src/meshlabplugins/filter_texture/filter_texture.cpp?revision=2258&view=markup
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> imran
>>>>
>>>>
>>>> I gave a very short look at the code.
>>>> A few suggestions (that in most cases holds for all my students too....)
>>>>
>>>> - keep the implementation of the filters out of the apply methods unless
>>>> it is a very short one (e.g. less than 10~20 lines of code). The ApplyFilter
>>>> function should be mostly a switch.
>>>> - try to structure your code a little more. Avoid very lengthy functions.
>>>>
>>>> - multiple atlases seems to me a bit a weird concept. The idea of atlases
>>>> is exactly the fact of having a single unique object. Let us assume that the
>>>> objective is an unique single atlas.
>>>> - You are doing a quite restructuring algorighm so it reasonable an
>>>> initial purging step. Perform at beginning a compactVertex and comapctface
>>>> call and get rid of all the IsD() tests in your loops.
>>>> - Do not take any care of number of multiplications, sums and operations.
>>>> Performance is affected by other things. Forget old style optimizations....
>>>> - stl has min and max, use them.
>>>> - pay attention to use int for truncating a value, floor and ceil make a
>>>> more predictable work, particualrly when negative numbers are around
>>>> - 2880 seems to me a quite weird value for a maximum texture size. Where
>>>> do it came from? why not a plain 4096 or 8192?
>>>> - on my machine, on the input mesh sample/texture_double.ply it does not
>>>> crash, but simply create trashed tex coords. From the log:
>>>>
>>>> created atlas of dimensions: 1024 , 512
>>>> inserted texture 0 into atlas at position: 0 0
>>>> inserted texture 1 into atlas at position: 512 0
>>>> saved texture atlas
>>>> original U: 0 original V: 0
>>>> new U: 2.82017 new V: 9300.4
>>>> original U: 1 original V: 0
>>>> new U: 2.82017 new V: 9300.4
>>>> original U: 1 original V: 1
>>>> new U: 2.82017 new V: 9300.41
>>>> original U: 0 original V: 0
>>>> ...
>>>>
>>>>
>>>>
>>>>
>>>> Cheers
>>>>
>>>> p.
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Paolo Cignoni -- Senior Researcher
>>>> Visual Computing Laboratory - ISTI - CNR http://vcg.isti.cnr.it/~cignoni <http://vcg.isti.cnr.it/%7Ecignoni> <http://vcg.isti.cnr.it/%7Ecignoni> <http://vcg.isti.cnr.it/%7Ecignoni>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> ISTI - CNR
>>>> Via Moruzzi 1,
>>>> 56124 Pisa
>>>> ITALY
>>>>
>>>>
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.net email is sponsored by: Microsoft
>>>> Defy all challenges. Microsoft(R) Visual Studio 2008.http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
>>>> _______________________________________________
>>>> Meshlab-devel mailing lis...@li...https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
>>>> Build the coolest Linux based applications with Moblin SDK & win great prizes
>>>> Grand prize is a trip for two to an Open Source event anywhere in the worldhttp://moblin-contest.org/redirect.php?banner_id=100&url=/
>>>>
>>>> ------------------------------
>>>>
>>>> _______________________________________________
>>>> Meshlab-devel mailing lis...@li...https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>>>
>>>>
>>>>
>>>> --
>>>> Paolo Cignoni -- Senior Researcher
>>>> Visual Computing Laboratory - ISTI - CNR http://vcg.isti.cnr.it/~cignoni <http://vcg.isti.cnr.it/%7Ecignoni> <http://vcg.isti.cnr.it/%7Ecignoni> <http://vcg.isti.cnr.it/%7Ecignoni>
>>>>
>>>>
>>>>
>>>>
>>>> ISTI - CNR
>>>> Via Moruzzi 1,
>>>> 56124 Pisa
>>>> ITALY
>>>>
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's
>>>> challenge
>>>> Build the coolest Linux based applications with Moblin SDK & win great
>>>> prizes
>>>> Grand prize is a trip for two to an Open Source event anywhere in the worldhttp://moblin-contest.org/redirect.php?banner_id=100&url=/
>>>> _______________________________________________
>>>> Meshlab-devel mailing lis...@li...https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>>>
>>>>
>>>> ------------------------------
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
>>>> Build the coolest Linux based applications with Moblin SDK & win great prizes
>>>> Grand prize is a trip for two to an Open Source event anywhere in the worldhttp://moblin-contest.org/redirect.php?banner_id=100&url=/
>>>>
>>>> ------------------------------
>>>>
>>>> _______________________________________________
>>>> Meshlab-devel mailing lis...@li...https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>>>
>>>>
>>>>
>>>> --
>>>> Paolo Cignoni -- Senior Researcher
>>>> Visual Computing Laboratory - ISTI - CNR http://vcg.isti.cnr.it/~cignoni <http://vcg.isti.cnr.it/%7Ecignoni>
>>>>
>>>> ISTI - CNR
>>>> Via Moruzzi 1,
>>>> 56124 Pisa
>>>> ITALY
>>>>
>>>>
>>>>
>>>> -------------------------------------------------------------------------
>>>> This SF.Net email is sponsored by the Moblin Your Move Developer's
>>>> challenge
>>>> Build the coolest Linux based applications with Moblin SDK & win great
>>>> prizes
>>>> Grand prize is a trip for two to an Open Source event anywhere in the
>>>> world
>>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>>>> _______________________________________________
>>>> Meshlab-devel mailing list
>>>> Mes...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>>>
>>>>
>>>
>>
>> -------------------------------------------------------------------------
>> This SF.Net email is sponsored by the Moblin Your Move Developer's
>> challenge
>> Build the coolest Linux based applications with Moblin SDK & win great
>> prizes
>> Grand prize is a trip for two to an Open Source event anywhere in the
>> world
>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>> _______________________________________________
>> Meshlab-devel mailing list
>> Mes...@li...
>> https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>>
>>
>
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's
> challenge
> Build the coolest Linux based applications with Moblin SDK & win great
> prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> _______________________________________________
> Meshlab-devel mailing list
> Mes...@li...
> https://lists.sourceforge.net/lists/listinfo/meshlab-devel
>
>
|
