CS 294-13: Assignment 2 — Real-Time / Image-Based RenderingRavi Ramamoorthi1 IntroductionIn this assignment, you will implement some of the modern techniques for real-time and/or image-basedrendering. To account for varied interests of students, I provide three choices for this assignment. You needonly pick one of the choices, based on your interests and expertise. Since many of these are close to theresearch frontier, I am providing as much guidance as possible, but you will need to be proactive in lookingup the original literature and other sources such as online tutorials. No source/skeleton code is available. Ifyou find online source code, you are welcome to use it as a reference, or to use utility functions, but obviouslynot as a core part of the assignment. If in doubt, please ask the instructor for clarification.This assignment should be completed in groups of two. Both partners will receive the same grade. Ifyou absolutely want to work alone, you may do so, but similar requirements will apply. There is currentlyno source code for this assignment, and as noted below, it will be turned in by creating a website. Futureeditions of the course may pr ovide more support, but the point is also to get you to work independently andcommunicate your results, given that this is an advanced graduate course.Choice of Assignment: There are three choices provided for this assignment, that correspond to threeimportant current research directions in rendering and appearance. You only need to do one of them (theyare listed in separate sections below).• Precomputed Radiance Transfer System: The technique of precomputed radiance transfer is a popularmethod over the last decade for rendering and relighting. By implementing a project along these lines,you will develop a basic background in real-time rendering in general, and in this new approach torelighting. I would suggest you start by implementing only the simplest basic system, such as Ng etal. 03, and if time permits, add to it. Students working on this project will have mathematical andsystems skills, with an interest in learning state of the art real-time rendering techniques.• Basic High Quality Rendering: Given the power of current graphics cards and computers, a number ofhigh quality visual effects can be included in the real-time rendering domain. This assignment asks youto explore many of these concepts. At a minimum, you should use GLSL shaders to implement basicshadow and environment mapping. You should also include some other visual effects. An ambitiousgoal would be include more general BRDFs, using something like Cabral et al.’s reflection-space image-based rendering. But even if you don’t do that, create some interesting visual effects and a really nicedemo of high quality real-time rendering. Students working on this project will have strengths inreal-time programming using OpenGL (or DirectX), and an interest in fast hardware rendering.• Image-Based Rendering: A popular new paradigm to rendering is to create new visual content fromexisting images by interpolating captured or synthetic data. In this project, you will write a basiclight field viewer, that will allow one to view real or synthetic data from various different angles. Notethe parallels to the real-time rendering projects above: precomputed radiance transfer for relightingessentially allows you to look at an object under any illumination, while this project allows you tolook at an object from different viewpoints. Both techniques are geared to r un in real-time. Studentsworking on this project will have interest in object and material acquisition, and stengths in developinginteractive s ystems.Given the nature of these methods, being close to research in the area, your best source is often theoriginal papers and online resources. You will have to do more groundwork on these projects, but it is also1a good preparation for the state of the art research in the area. Choose which of the three projects aboveyou want, based on your group’s strengths and interests. They are described in more detail below.2 Submitting the Assignment and LogisticsSince one of the goals of preparing for research is to also communicate clearly, the assignment must be turnedin either by submitting a PDF by e-mail to the instructor (as you would a scientific paper), or by creatinga website and sending a pointer to it. Since part of your results will involve interactive effects, we will alsoschedule individual demonstrations with groups to demo your system in action. If you want, you can alsocreate a video from screen-captures of your system, available from your website (as you would for a paper),but this is not required.Document your results properly, showing the various types of scenes your method can handle. Explainany interesting algorithmic aspects. It is likely you will need to create some test scenes; there are a numberof resources available on the web for interesting geometric models and scenes that have been used in previouspapers.Failure to fully document the results of the program will likely lead to unhappiness from me. Be honest,and note what works and what doesn’t; this will be the most fun, and make the assignment easiest to grade.Some part of the grade will be reserved for the clarity of the writeup.If you make a website, do not modify it after the due date. In either case, please include links to thesource code and executable, and again do not modify them after the due date.Support: Feel free to contact me regarding questions. You may also use the course newsgroup. There isno code framework provided for this assignment. You are welcome (and encouraged) to look at online codesources and the original literature, and even the code as a reference, or to create utility functions. Obviously,you cannot copy code from elsewhere for a core part of the assignment though; ask me if there are anyquestions.3 Precomputed Radiance Transfer and RelightingRecently, a body of work known as precomputed radiance transfer, or precomputation-based relighting, hasbecome a popular technique for real-time rendering in varying illumination, and recently also viewpoint fordynamic scenes. Good references are the seminal papers by Sloan, Kautz and Snyder in SIGGRAPH 02,and Ng, Ramamoorthi, Hanrahan in SIGGRAPH 03. A good historical source, that also surveys some of theprevious work in image relighting, is a survey I recently wrote on precomputation-based rendering (availableon both my
View Full Document
Unlocking...