This is the first version of my vector field design system for the class CS550 (Intro to Computer Graphics) by Dr Ron Metoyer. It is only a toy program and contains some bugs, but I will not update it any further. The reason I post it here is because this is my first graphics project :). There is nothing complicated here. Just an implementation of IBFV and a simple interface to allow you to add some design elements in the field. Interesting people can find a more complete design tool of mine here.

This is the final project for the class CS551 (Advanced Computer Graphics)  by Dr Ron Metoyer. In this project, I have implemented the classical distributed ray tracing (project webpage) to produce the effects including soft shadow, rough refraction, depth of field and motion blur. More details can be found in the project webpage (including source code). Generally, I was pretty satisfied with the results I got, but for complicated scene the performance of the program was not that good.

The images to the left were generated by the program implemented for the class CS519 Advanced topic: Geometry Processing by Dr. Eugene Zhang. The basic idea is to compute the gradient and edge fields of the original images, and then use the rectangle strokes to generate the painting effects. We also implemented other NPR algorithms, such as toon rendering and interactive silhouette extraction of a 3D object.

The images to the left display some course work for the class CS519 Shader and GLSL by Dr. Mike Bailey. The shader Olympic finally became the poster in Siggraph 2006.

The images to the left will be the selected from the course work of the class CS554 Geometric Modeling by Dr. Eugene Zhang

The final project for the class CS534 Machine Learning by Dr. Alan Fern

[Abstract] Vector field design is critical for many visualization and computer graphics tasks, such as vector field visualization, fluid simulation, texture synthesis, 3D surface manipulation, animation synthesis, and more. In this project, we have made the first trial of time-dependent vector field design. To do that, we employ the idea of keyframe design widely used in computer animation community. That is, we design the time-dependent vector field sequences by specifying only a few keyframe slices, and the system will generate the complete sequence of desired vector field by interpolating the keyframe slices using particular function (path). In order to determine a proper interpolation function, we make use of the machine learning techniques. Specifically, we first propose a machine learning model for time dependent vector fields. We then simplify the model and constrain ourselves to a smaller function space. Following, we use a simple gradient descent like learning to extract the proper interpolation function from the given examples. We finally demonstrate the utility of machine learning techniques in helping to find the interpolation function by providing a number of design results.

[pdf] of the final report

[Abstract] In this project, I implemented the work of Continuum Crowds by Treuille et al. [2006]. I have emphasized the importance and difficulties of crowd simulation in many applications in the project proposal. The goal I pursue in this project is a crowd generation tool with reasonable complexity and interactive control. Without sacrificing the reality of the obtained crowd effects, continuum crowds presents a promising approach to achieve that goal.

The functionalities being implemented in this project include a crowd simulation framework based on the unit cost model proposed in [1], a user interface to allow the user the ability to modify the environment and the goals during run time, an extension to the explicit path control through flow design and the ability of naive 3D visualization.