Point Cloud in Nuke using "PositionToPoints", 3delight Rendman Shader Language and cortex-vfx

Here is the end result:

First things first I need a model to work with

Model

• Stanford Bunny OBJ via MD Fischer

Shader

surface bakeColourAndPosition(
uniform float diffuseAmount = 1;
varying color surfaceColour = color(0.18,0.18,0.18);
varying color opacityColour = color(0.99,0.99,0.99);
uniform string bakeFile="/tmp/out.bake";
)
{
   varying normal Nn = normalize(N);
   Ci = diffuse(Nn)*surfaceColour*diffuseAmount*Cs;
   Oi = opacityColour*Os;
   varying point Pworld = transform("current","world",P);
   bake(concat(bakeFile,"Position"),s,t,Pworld);
   bake(concat(bakeFile,"Colour"),s,t,Ci);
   Ci *= Oi;
}

This shader will produce two text "bakefile" files in the /tmp directory

Note: The texture coordinates are ignored, only using the 3rd,4th and 5th values

1. one named out.bakeColour with colour information
2. one named out.bakePosition with position information

Output Bakefiles

• Download .tar.gz here...

  Due to the SIMD nature of shaders the line count of each of the bake files is the same so they should contain line for line the same information about the Position and Colour Respectively.

  They are in ASCII format so they are easy enough to parse with Python.

  here is an example of thier content:

  
  out.bakePositionmh
  3
  0 1 -0.07415867 0.17987273 -0.05079475
  0 1 -0.073529155 0.1800126 -0.051191031
  0 1 -0.07289961 0.18015243 -0.051587344
  0 1 -0.072270096 0.18029229 -0.051983685
  0 1 -0.07164058 0.18043211 -0.052379965
  0 1 -0.07101102 0.18057197 -0.052776248
  0 1 -0.07038155 0.1807118 -0.053172619
  

  Creating Position and Colour EXR files using cortex-vfx

  
  #!/usr/bin/env python
  
  import sys,os,math
  
  IECoreInstallPath = "/usr/lib/python2.6/site-packages"
  
  if IECoreInstallPath not in sys.path:
  	sys.path.append(IECoreInstallPath)
  
  from IECore import *
  
  bakeFolder = "/tmp"
  
  colorBakeFileLocation = os.path.sep.join([bakeFolder,"out.bakeColour"])
  
  positionBakeFileLocation = os.path.sep.join([bakeFolder,"out.bakePosition"])
  
  
  
  
  def parseBakeFile(bakeFileLocation):
  	data = []
  	counter = 0
  	bakeFile = open(bakeFileLocation,"r")
  	for line in bakeFile.readlines():
  		counter +=1
  		if counter > 2:
  			stuff = line.strip().split(" ")
  			if len(stuff) > 2:
  				data.append((float(stuff[2]),float(stuff[3]),float(stuff[4])))
  	print "Completed parsing %d lines of file %s" % (len(data),bakeFileLocation)
  	bakeFile.close()
  	return data
  
  colourData = parseBakeFile(colorBakeFileLocation)
  
  positionData = parseBakeFile(positionBakeFileLocation)
  
  if len(colourData) == len(positionData):
  	squareSize = int(math.sqrt(len(positionData))) +1
  	print "Square Size: %d, Excess Pixels : %d" % (squareSize,squareSize*squareSize - len(colourData))
  	width = squareSize
  	height = squareSize
  	x = FloatVectorData( width * height )
  	y = FloatVectorData( width * height )
  	z = FloatVectorData( width * height )
  	
  	r = FloatVectorData( width * height )
  	g = FloatVectorData( width * height )
  	b = FloatVectorData( width * height )
  	
  	for i in range(len(colourData)):
  		r[i]=colourData[i][0]
  		g[i]=colourData[i][1]
  		b[i]=colourData[i][2]
  		x[i]=positionData[i][0]
  		y[i]=positionData[i][1]
  		z[i]=positionData[i][2]
  		
  	boxColour = Box2i( V2i( 0, 0 ), V2i( width-1, height-1 ) )
  	boxPosition = Box2i( V2i( 0, 0 ), V2i( width-1, height-1 ) )
  	
  	imageColour = ImagePrimitive( boxColour, boxColour )
  	imagePosition = ImagePrimitive( boxPosition, boxPosition )
  	
  	imagePosition["R"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, x)
  	imagePosition["G"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, y)
  	imagePosition["B"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, z)
  
  	imageColour["R"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, r)
  	imageColour["G"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, g)
  	imageColour["B"]= PrimitiveVariable( PrimitiveVariable.Interpolation.Vertex, b)
  	
  	writePosition = Writer.create( imagePosition, "/tmp/outPosition.exr" )
  	writeColour = Writer.create( imageColour, "/tmp/outColour.exr" )
  	writePosition.write()
  	writeColour.write()
  

  See more about cortex-vfx on google code:

  • http://code.google.com/p/cortex-vfx/

  Using Nuke to read the Position and Colour Data

  
  

  If you werent able to create your own pair of EXRs you can download the pair here in .tar.bz2 format

  So then you just need to connect them up to the input nodes for the PositionToPoints 3d node as follows:

  
  

  If you thought this was useful leave a comment, or if you thought it was stupid leave a comment about how to improve it

  Sam