Homework 0: Transformations & OpenGL Warmup
The goal of this warmup assignment is to get comfortable with the
programming environment you will be using for this class and
familiarize yourself with a simple library that we will use for linear
algebra. It's also an opportunity for a crash course in C++ and
OpenGL (if you're not already familiar with them). This
assignment is only worth 1/4 of the points of a regular homework
assignment. It will not be rigorously graded, but you are expected to
work through and submit the exercises. Please try to finish this
assignment by the end of the first week of classes; however, there is
no firm deadline for this assignment.
Here are a couple tutorials you may want to check out:
The incidental goal is also to have fun with bizarre fractal objects.
IFS are self-similar fractals: a subpart of the object is similar to
the whole. The classic example of an IFS is Barnsley's fern, where
each subpart of the fern is exactly the same as the whole fern. IFS
are described by a set of affine transformations (rotations,
translations, scale, skew, etc.) These transformations capture the
self-similarity of the object. IFS can be defined in any dimension,
and we will play with both two-dimensional and three-dimensional ones.
Formally, an IFS is defined by n affine transformations. Each
transformation fi must be contractive: The
distance between points must be reduced. An attractor of the
IFS is the object such that A = U fi (A).
A is unchanged by the set of transformations: It is a fixed
We can render an IFS by iterating the transform on random input points
from the unit square. We approximate the fixed point by applying the
transformation many times. The algorithm is as follows:
for "lots" of random points (x0, y0)
for k=0 to num_iters
pick a random transform fi
(xk+1, yk+1) = fi(xk, yk)
display a dot at (xk, yk)
To reduce the number of points necessary to make an image of
reasonable quality, probabilities are assigned to each transformation,
instead of choosing a transformation with uniform probability.
- Download the provided source code and set up your C++ development
environment. To receive full credit, your homework assignments must
compile and run without errors using gcc/g++ on a Linux/MacOSX system.
Even if you plan to do much of your development in another
environment, you'll probably still want to set up gcc so that you can
test it before submission. For interactive display of your IFS, you
will use the OpenGL API that uses graphics hardware for fast rendering
of 3D primitives.
A CMakeLists.txt file is provided and has been tested on
Linux(gcc), MacOSX(gcc), and Windows(Visual Studio). Please see the
TA or instructor in office hours if you have trouble setting up your
All files implementing OpenGL code should include the OpenGL
header files (gl.h, glu.h, glut.h), and some platforms need additional
commands. Please copy the full block of gl include syntax from
glCanvas.h, to make your code as portable as possible.
The initial executable from the provided code should launch an
OpenGL window and draw a solid cube or a cube of noisy points and you
should be able to navigate the scene with the mouse (left button
rotates, middle button translates, shift+mouse zooms). Try these
- Now you're ready to start coding. Write a C++ class
IFS that renders iterated function systems, including the
class declaration (in the header file ifs.h) and the
implementation (ifs.C). The IFS class should include:
- a field to store n, the number of transformations,
- an array/vector of matrices representing the n transformations,
- an array/vector of the corresponding probabilities for choosing a
- a draw method which makes appropriate OpenGL calls to draw points
or polygons to the OpenGL window, and
- When you're done, adjust the main function (provided in
main.C) and other code, as necessary.
- Use the linear algebra library for the point and transformation
- Consider the performance of your programming environment.
How many polygons/points can you render interactively? What
improvements could you make to your code?
A small amount of extra credit is available for creative
extensions. Examples include:
Include a short paragraph in your README.txt file describing
- design a new IFS --- figure out the transformations and probabilities,
- implement an interesting, non trivial color scheme,
- implement anti-aliasing,
- experiment with depth-first vs. breadth-first, etc.
- The (included)
MersenneTwister random number generator can be used to generate
random integers and random floating point numbers.
- To debug your code, set the number of iterations to one. This
will allow you to check that you got the transformations right.
- Be careful, arrays are indexed from 0 to n-1 in
C++. Reading beyond the bounds of the array will probably result in a
- Use assert() to check function pre-conditions, array
- To perform transformations in OpenGL, read about the Modelview
matrix stack and the OpenGL commands glMatrixMode(),
glPushMatrix(), glPopMatrix(), and
glMultMatrix() in the
OpenGL Programming Guide, Chapter 3.
- Linear Algebra Library (vectors.h & matrix.h & matrix.cpp)
Linear algebra support for floating point vectors with 3 and 4
elements (Vec3f and Vec4f) and 4x4 floating point
matrices (Matrix). For this assignment, the void
Matrix::Transform(Vec3f &v) function will be handy.
- Parsing code for command-line arguments and input files
The program takes a number of command line arguments to specify the
input file (-input), number of points (-points),
number of iterations (-iters), and window height & width
(-size). Your program should render points by default, or
cubes if -cubes is specified. Examples are shown below.
Code to parse input files and command line arguments is provided.
- OpenGL and main code (main.cpp, ifs.h, ifs.cpp, glCanvas.h,
glCanvas.cpp, camera.h, camera.cpp)
OpenGL programs can be tricky to set up from scratch. This base code
should do all that work for you.
- Data files (fern.txt, dragon.txt, sierpinski_triangle.txt, and giant_x.txt)
The input data for an IFS is a file which contains n,
the number of transforms, followed by the probability of choosing each
transform and a 4x4 floating point matrix representation of the
./ifs -input sierpinski_triangle.txt -points 10000 -iters 0 -size 200
./ifs -input sierpinski_triangle.txt -points 10000 -iters 1 -size 200
./ifs -input sierpinski_triangle.txt -points 10000 -iters 2 -size 200
./ifs -input sierpinski_triangle.txt -points 10000 -iters 3 -size 200
./ifs -input sierpinski_triangle.txt -points 10000 -iters 4 -size 200
./ifs -input sierpinski_triangle.txt -points 10000 -iters 30 -size 200
./ifs -input fern.txt -points 50000 -iters 30 -size 400
./ifs -input giant_x.txt -points 10000 -size 400 -iters 0
./ifs -input giant_x.txt -points 10000 -size 400 -iters 1
./ifs -input giant_x.txt -points 10000 -size 400 -iters 2
./ifs -input giant_x.txt -points 10000 -size 400 -iters 3
./ifs -input giant_x.txt -points 10000 -size 400 -iters 4
Now here's an example using the -cubes argument. Each cube
represents the bounding box of the cloud of random 3D points that are
transformed in each iteration. There are a couple different ways to
implement this rendering mode.
./ifs -input giant_x.txt -size 400 -iters 0 -cubes
./ifs -input giant_x.txt -size 400 -iters 1 -cubes
./ifs -input giant_x.txt -size 400 -iters 2 -cubes
./ifs -input giant_x.txt -size 400 -iters 3 -cubes
./ifs -input giant_x.txt -size 400 -iters 4 -cubes
Please read the Homework information page again before submitting.