Forum

Getting emguCV to run

Hey @circuitb

Your results sound right for small triangles. I’ll be looking at this today
I chose to normalise the area (since position output is normalised in VVVV)
I.e. Space is -1 to +1,

something 10px wide at 640x480 is about 1/50th the width and height in normalised coords
= 2/50 in normalised coords
= 0.04 of the width and height
Therefore area = 0.04 * 0.04 = 0.0016 = 1.6E-3

If you have something that’s a single pixel (I.e. Noise)
Then you have 0.000016 = 1.6E-5
This is the area of 1px*1px at VGA in normalised coordinates
This would also be a common area for a triangle at the perimeter of an object

If you increase the input resolution to HD
1px is about 2E-6

the minimum area module gives me the correct results and uses contour’s area output
But havent really looked at Delaunay area output

Can you check again if these results make sense?

Contour’s under development (added some nice polynomial approximation, and playing with the delaunay node a lot).
But not got Internet here so can’t push the changes for a couple of days.

Elliot

@circuitb - will look at errors, but I guess that must be Line (Ex9.Geometry) bugging out since it’s kind of impossible to send it corrupt data in vvvv

I presume it just doesn’t like the constantly changing high spread counts.

I’m finding that VertexBuffer/Mesh also doesn’t like constantly changing high spread counts (converting Delaunay into mesh)
This is a bit weird since i’ve done this in OpenGL before without any issue, and I’m sure DX must be the same.
Geometry shaders would help of course…

Elliot

looking for the next compiled alpha release since i dont have visual studio…

I’ve made lots of changes whilst here at Cynenet festival in Desden
Will post dll’s when the pc is back on an internet connection when I get back to Seoul