Bernard Llanos
- A MATLAB script for generating point clouds, where the points are arranged along curves, describing superellipsoids
- An OpenGL application which reads curve-structured point cloud files incrementally,
and estimates normal vectors at point positions, using several different
algorithms
- Ignoring normal vector estimation capabilities, this program can be used to view point clouds produced by the 3D probing project (https://github.com/bllanos/linear-probe).
- A MATLAB script for quantitative evaluation of normal vector estimation on the curve-structured point clouds
- This program will not run in the CSC 1-59 lab environment, because it requires additional libraries.
- MATLAB is required to run the data generation and evaluation components of the project. MATLAB 2017b was used for development, but a much older version of MATLAB should be sufficient. No additional MathWorks toolboxes should be required.
- Development was carried out on Linux (Ubuntu 16.04, for instance).
- Computational Geometry Algorithms Library (Ubuntu packages 'libcgal*'),
version 4.11-1 (https://www.cgal.org/)
- Depends on the GMP and Boost libraries (https://doc.cgal.org/latest/Manual/installation.html)
- Eigen3 (Ubuntu package 'libeigen3-dev'), version 3.3 (http://eigen.tuxfamily.org/index.php?title=Main_Page)
- FreeGLUT (Ubuntu packages 'freeglut3*'), version 2.8.1 (http://freeglut.sourceforge.net/)
- OpenGL 2.1 or higher
- OpenGL Extension Wrangler Library (Ubuntu packages 'libglew*'), version 1.13 (http://glew.sourceforge.net/)
- Run
make pointCloudViewerto produce the executable 'pointCloudViewer'.
- Run './MATLAB/SampleSuperellipsoids.m' to generate point cloud data files, which presently will be saved in './superellipsoids/'.
./pointCloudViewer $FILENAME, where 'FILENAME' is the path to one of the point cloud data files. A graphical application will launch and start building the point cloud on screen.- Press 'g' and 'G' to cycle back and forth between normal estimation algorithms. Some algorithms are only implemented in batch form, and will update all points when they are activated. As the point cloud continues to be built, these batch algorithms are not run except when they are re-selected using the 'g' and 'G' keys.
- Press 'e' or 'E' to save the estimated normal vectors at points in the point cloud to the file './results.csv'. Results are saved for all normal vector estimation algorithms at once.
- A full list of keyboard controls is provided below.
- Run './MATLAB/Evaluation.m' to summarize the errors of the normal vector estimation algorithms. The script presently performs its calculations on the data in './results.csv'.
- The curve of points which is currently being read is displayed in cyan.
- Normal vectors are implicitly represented by lighting the scene. A single point light source is to the right of, and above, the camera.
- Existing curves are displayed in green to yellow if their normal vectors deviate from the reference normal vectors by zero to 10 degrees, respectively.
- Existing curves are displayed in yellow to red if their normal vectors deviate from the reference normal vectors by 10 to 45 (or greater) degrees, respectively.
A CSV file with the following columns:
- Curve index: The actual value is ignored, but a change in values between lines results in a new curves being created
- Point index along the curve (ignored)
- x-coordinate of the point position
- y-coordinate of the point position
- z-coordinate of the point position
- x-component of the point's normal vector hypothesis
- y-component of the point's normal vector hypothesis
- z-component of the point's normal vector hypothesis
The normal vector hypothesis is a unit vector within 90 degrees of the true normal vector.
Identical to the input file format, but with the following three additional columns:
- x-component of the curve tangent vector at the point
- y-component of the curve tangent vector at the point
- z-component of the curve tangent vector at the point
The normal vectors in the output files represent estimated normal vectors, as opposed to normal vector hypotheses.
- 'a'/'A': Rotate the camera around its local Y-axis, in either direction, respectively
- 'b': Decrease width of rasterized lines
- 'B': Increase width of rasterized lines
- 'c'/'C': Translate the camera along its -Y and +Y directions, respectively
- 'd'/'D': Translate the camera along its -X and +X directions, respectively
- 'e'/'E': Write normal vectors, point positions, and estimated point tangent vectors to a file for use in quantitative evaluation.
- 'f': Turn off fog effect
- 'F': Turn on fog effect (default)
- 'g'/'G': Cycle back and forth between the different normal estimation algorithms. There are 6 algorithms available, including simply using the reference normal vectors.
- 'l'/'L': Rotate the camera around its local Z-axis, in either direction, respectively
- 'n'/'N': Translate the point cloud along its -Z and +Z directions, respectively
- 'p': Resume building the point cloud from the input file
- 'P': Pause building the point cloud from the input file
- 'q': Quit
- 'r'/'R': Rotate the point cloud around its local Z-axis, in either direction, respectively
- 's': Decrease size of rasterized points
- 'S': Increase size of rasterized points
- 't'/'T': Rotate the camera around its local X-axis, in either direction, respectively
- 'u'/'U': Rotate the point cloud around its local X-axis, in either direction, respectively
- 'v': Switch to orthographic projection
- 'V': Switch to the default perspective projection
- 'w'/'W': Write the point cloud data out for debugging purposes. The output file, './out.csv', is the same as the input file, except that the lines starting with 'vn' and '#t' are the estimated surface normal vectors and curve tangent vectors, respectively.
- 'x': Reset camera pose, clear the point cloud, and pause. When the program is resumed, the point cloud file will be read again starting from the beginning.
- 'y'/'Y': Rotate the point cloud around its local Y-axis, in either direction, respectively
- 'z'/'Z': Translate the camera along its -Z and +Z directions, respectively
- '-': Decrease the rate at which the point cloud is built
- '+': Increase the rate at which the point cloud is built
- 'DOWN'/'UP': Translate the point cloud along its -Y and +Y directions, respectively
- 'LEFT'/'RIGHT': Translate the point cloud along its -X and +X directions, respectively
Each header file lists references consulted when creating it and the corresponding '.cpp' file (if there is one).
I developed the project code and my code for Assignments 1-3 together.