Posts Tagged ‘Renderman’

trying to calculate localVisibility with Spherical Harmonics

Saturday, November 12th, 2011
Spherical Harmonics Coefficients of Local Visibility

Spherical Harmonics Coefficients of Local Visibility

In hidsight, baking a lookup table of vectors and the spherical harmonics, might lead to artifacts, but with 1024 samples it doesnt look too bad

Feel free to download the source to sing along, its all based on the Sony Paper from 2003 see the PDF from SCEA

Python executable that write Spherical Harmonics Renderman header: sh1.py

Please find the resulting header file attached: SH.h

And a simple shader to calculate transmission based on the table of stratified samples above: localVisibility.sl

Its a bit clunky because I couldn’t work out how to do two dimensional arrays in RSL, im glad to fix it up if it is possible.

The preprocessor thing isnt that sweet either :/

As below:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
#include "SH.h"
 
surface localVisibility(
        uniform float maxDistance = 10; 
        uniform string outputFolder = "";
)
{
        SHVECTOR
        SHSPH0
        SHSPH1
        SHCOEFF0
        SHCOEFF1
        SHCOEFF2
        SHCOEFF3
        SHCOEFF4
        SHCOEFF5
        SHCOEFF6
        SHCOEFF7
        SHCOEFF8
        vector Nworld = vector(transform("world",N));
        point Pworld = transform("world",P);
        uniform float numSamples = 1024;
        uniform float numCoeffs = 9;
        varying float results[9]={0,0,0,0,0,0,0,0,0};
        uniform float i,j;
        varying float faceforward = 0;
        varying float occl = 0;
        for(i=0;i<numSamples;i=i+1){
                float Hs = samplesVector[i].Nworld;
                point destinationWorld = Pworld + samplesVector[i]*maxDistance;
                point destinationCurrent = transform("world","current",destinationWorld);
                if (Hs > 0){
                        faceforward += 1;
                        float isHit = comp(transmission(P,destinationCurrent),0);
                        if (isHit > 0)
                        {
                                occl += 1;
                                for(j=0;j<numCoeffs;j=j+1){
                                        varying float value = isHit;
                                        if (j == 0)
                                        {
                                                value *= samplesCoeffs0[i];
                                        }
                                        if (j == 1)
                                        {
                                                value *= samplesCoeffs1[i];
                                        }
                                        if (j == 2)
                                        {
                                                value  *= samplesCoeffs2[i];
                                        }
                                        if (j == 3)
                                        {
                                                value *= samplesCoeffs3[i];
                                        }
                                        if (j == 4)
                                        {
                                                value *= samplesCoeffs4[i];
                                        }
                                        if (j == 5)
                                        {
                                                value *= samplesCoeffs5[i];
                                        }
                                        if (j == 6)
                                        {
                                                value *= samplesCoeffs6[i];
                                        }
                                        if (j == 7)
                                        {
                                                value *= samplesCoeffs7[i];
                                        }
                                        if (j == 8)
                                        {
                                                value *= samplesCoeffs8[i];
                                        }
                                        results[j] += value;
 
                                        }       
                                }
                        } 
                }
        for (j=0;j<numCoeffs;j=j+1){    
                results[j] /= faceforward;
                }
        occl /= faceforward;
        faceforward /= numSamples;
        Ci = color(results[0],results[1],results[2]);
        Oi = 1;
        Ci *= Oi;
 
}

I dont think it is working yet, but it compiles and renders

Sam

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

Thursday, October 14th, 2010

Here is the end result:

Nuke Node "PositionToPoints" with 3d EXR inputs

First things first I need a model to work with

Model

Shader

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
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:

    1
    2
    3
    4
    5
    6
    7
    8
    9
    
    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

    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14
    15
    16
    17
    18
    19
    20
    21
    22
    23
    24
    25
    26
    27
    28
    29
    30
    31
    32
    33
    34
    35
    36
    37
    38
    39
    40
    41
    42
    43
    44
    45
    46
    47
    48
    49
    50
    51
    52
    53
    54
    55
    56
    57
    58
    59
    60
    61
    62
    63
    64
    65
    66
    67
    68
    69
    70
    71
    72
    73
    74
    75
    76
    77
    
    #!/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:

    Using Nuke to read the Position and Colour Data

    File > Script Command [X] > PositionToPoints

    File > Script Command > PositionToPoints

    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:

    Nuke Node "PositionToPoints" with 3d EXR inputs

    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