Yeah! Looks much better now :)
Fabrice
On Nov 16, 2008, at 3:51 PM, corin_w wrote:
Hi everyone
thanks very much everybody especially Fabrice, I have got it working.
it was me giving 0-image width or height, u, v coordinates.
Please take a look at the now properly textured 3d face.
I will post the code up as soon as I've ironed out a few bugs
cheers, Corin.
On Nov 15, 11:19 pm, Fabrice <[EMAIL PROTECTED]> wrote:
the uvs must be zero/one based, thats why you get that offset
u = 1/width image
v = 1- 1/height image
but on top of my head
if you have a for instance if you work by colums a sort of a quadface
information
like
a b
c d
then face is cab then cbd
if the face is inversed
do then acb and bcd
look at one of the extrude classes, especially where vlue flip is
used, there are also details on the face order
for the uvs
Hope it helps
Fabrice
On Nov 15, 2008, at 11:40 PM, corin_w wrote:
Basically one of the guys from littlebigplanet released some source
code for the scanner that he had made as an experiment, im trying to
port this over to flash.
the scan is made from 3 photos of the subject with stripes projected
onto it. the stripes are sinusoidal so when you add all of them
together you get a plain black and white photo of the subject.
this is
the texture.
basically the code scans across each pixel in the 3 input images and
measures the average difference between that pixel and its
neighbours
and so can tell where in a projected stripe that pixel lies from the
distortion of this by the uneven surface on to which the line
pattern
is projected it is possible to triangulate the depth of that pixel.
the part of the code that creates the mesh simply steps through the
pixels of the image column by column and generates a vertex at that
x,y value with z being the triangulated value.
u and v are also the x y coordinates.
if I make the size of the triangles very large and put a vertex a
step
of every 50 pixels it is possible to see the border between the
yellow
and blue squares on the texture moving up and down the faces as if
there is some error in the way that the textures are being
transformed.
what values should u and v take? x and y of the image or something
weird?
On Nov 15, 7:11 pm, Fabrice <[EMAIL PROTECTED]> wrote:
Can you pass in the object int {material:mat, debug:true}
so that we can see the faces?
Even if faces would be overlapping, this mesh do not have the
propper
face order and have wrong uv mapping.
Is it possible that your scan is considering a certain volume in
space
and that the uvs values are based on this? Most info is coming from
the
top of the image, or is it doing a 180 degree scan on the x axis?
Looking at the the first model showing with no dought a face, I'm
100%
positive that if you arrange the vertexes per row and pass it to
the
Skin class
that it would work. However if your vertex lists are too unequal to
feed the class then I would consider project from a plan x/y and
define the uv's from the vertexes.
But then the uv's would no longer respecting the scan.
Can you extend a bit on the " 3d scan of a face ported to as3 from
Processing"
What is exactly coming out Processing? How is the convertion to
Away
done?
Can you paste one or two lines of a face definition generated by
Processing?
Fabrice
On Nov 15, 2008, at 6:38 PM, Peter Kapelyan wrote:
Hi,
For some reason it looks as though there are two sets of faces on
top of each other, or basically two of the exact same model
overlapping each other, when I zoom in i see some faces go from
green to red (the same "triangles").
That happens when two of the exact same object are both in the
scene
in the same place.
Using Fabrices test, you see your object isnt using the whole
texture.
Is there some way you can test this with lower resolution data? By
using less info you might be able to see exactly what is going
wrong, I had to zoom in to tryt o see what may be going on.
If you tried something with say only 10 triangles that use the
whole
texture, it might be easier to pinpoint the problem(s).
By the way, I read about this experiment that you started several
months back, last week. I would love to see you make it happen in
Flash, it is an awesome idea!
-Pete
On Sat, Nov 15, 2008 at 11:36 AM, corin_w
<[EMAIL PROTECTED]> wrote:
Hi,
http://corinwilkins.co.uk/face.html
uploaded a new version with a very simple background strange
goings on
still ongoing
will updating my version of away 3d beable to fix this?
On Nov 15, 1:21 pm, Fabrice <[EMAIL PROTECTED]> wrote:
why dont you try a simple BitmapMaterial just to be sure its
not a
mapping issue?
To make a quick check of the uvs, just make a texture divided
with 4
colored rects, each a 1/4 of the source rect, then center,
where 4
squares meet
should be located at +- the "nose" of the model if the uvs are
right.
And I see you use an old version of the engine, 1.9.4, you should
run
the latest trunk, many fixes and enhancements
were made since that release.
Fabrice
On Nov 15, 2008, at 1:59 PM, corin_w wrote:
Hi,
http://corinwilkins.co.uk/face.html
this is a 3d scan of a face ported to as3 from processing, on
the left
is the texture that should be mapped onto the face, but as you
can see
something isnt quite right.
the uv coordinates are almost certainly correct i can only think
its a
lighting issue as there is a flicker on the chin of the face
at a
certain rotation.
any ideas?
cheers, corin
--
___________________
Actionscript 3.0 Flash 3D Graphics Engine
HTTP://AWAY3D.COM
