CS559 Course News

I will be out of town Thursday November 18th and office hours will (therefore) be cancelled.

We will have a guest lecturer in class to talk about realistic rendering.

Project 3 is now "officially" announced.

A preview of project 3 has been posted (look here.)

It is unlikely that the project description will change much, but it might.

For different features that have multiple parts, sometimes the parts do add up.

• Having both rail ties and parallel rails is worth more than having either one, but probably not as much as having both (if it were really amazing you could get all 17 points, but it would have to be amazing). Just drawing a line for the rails and drawing lines for the rail ties is a simple way to do "fancy tracks" (better is to draw polygons to give the them some thickness).
• The other "multiple parts of one things" vary. simple physics would be 5 points for momentum, and 5 more points for having the "ramp up".
• for multiple cars, 10 points is for having multiple cars on the track at the same time with the correct spacing. 5 additional points (for a total of 15) if the cars are different designs.

You should also be aware that the point values are for doing things correctly. You will get some subset of the points for doing simpler variants.

This is a bigger mistake that you may run into should you implement cubic B-Splines. The matrix on page 340 is upside down (e.g. the bottom row should be the top).

Everywhere in the chapter, I use the convention that the u vector is [u^0 u^1 u^2 ... u^d]. Except here, so it should be fixed.

We have decided to change around the project and assignment due dates.

Projects 2 and 3 have been delayed 1 week. Project 2 is now due on November 15th, and Project 3 will be due on December 13th.

Note that the due dates of the written homework assignments have been moved up.

Yu-Chi will email everyone their project 1 grades. Basically, from your demo scoresheet you can figure out what grade to expect with a slight penalty if you didn't follow directions.

If you get your grade and it is different than what you expected, we will explain where it came from. If you think we've made an error, we will regrade your project, which may lead to a different result.

Similarly, if you think your exam has been misgraded, please let us know and we'll regrade it.

The answer key to the midterm is posted here.

Yu-Chi will email everyone with their score and grade. If you'd like to see your exam, please stop by his office.

There will be a graphics seminar on Monday, October 31st at 4pm in 4310 CS that may be of interest. The speaker is Pradeep Sen from Stanford University. He will be speaking on "Dual Photography" - the topic of his recent SIGGRAPH paper. I expect that the talk will be accessible to 559 students.

Yes, the midterm was hard. Everyone thought so.

I am aware that there are some questions that were either incorrect or ambiguous.

I am also aware that the exam took much longer than I had thought. This is a problem since many people had to leave promptly at 10:45 to get to other classes (and at least one person had to go to another exam - YIKES!).

I haven't figured out what to do about it yet. Given all the effort that went into designing and checking the exam (yes, there were beta-testers, and at least one of the bugs in th exam was introduced in response to their comments), you would have thought I would have realized that there was going to be a problem.

For the 4 people who did not show up for the exam, please confirm that you have dropped.

Were put on the web but not linked to (oops!). They are here.

At least one student has discovered the tutorials at http://www.xmission.com/~nate/tutors.html and found them useful. I will try to get them installed on the CSL machines so that you can play with them.

Written homework 3 has been posted here. Its not due until after the next project, however, the first 5/7ths of it may be good practice for the exam.

With Bezier curves, the first derivatives are obtained by multiplying the vector between the first two points by the degree of the curve. I have confirmed this in 2 other textbooks, as well as by deriving the algebra for the 2nd order case.

So, the CAPTION of Figure 15.10 is incorrect (and the literal scaling of the little vectors that symbolize derivatives). The equations are correct, and the text is correct.

This figure was added late in the game (it is not in the original draft that I sent to the publisher). Still, I should have caught this in proofreading.

(Update. I was told that this mistake was introduced at the last minute after my last chance to proofread, so there was no way for me to catch it.)

I have made an outline of all the topics that we have covered in class and placed it here.

I have also fixed the calendar to better reflect what we actually covered, and to make sure that the readings actually correspond to the lectures.

As many of you noticed, project 2 is posted. The links to it should now actually work.

The Midterm Exam will be held in B371 Chemistry on Tuesday 10/18 at 9:30am (it is the normal class time). The exam will be designed to take 1 hour, but you will have the full 75 minutes. The exam is closed book/closed notes and you should not need a calculator (you'll have to do some basic arithmetic).

All topics discussed in class (and their associated readings) are fair game. In particular, there will be questions on curves. However, there will not be questions about B-Splines (Section 15.6 second half), not are you responsible for the derivations of cubic basis functions (you'll find out about these soon enough).

I expect to post a "summary" of the topics that we've discussed in class (and therefore may appear on the exam), as well as some practice problems later this week.

Q1: Please notice the equation of the convolution equation, h(t_0) = integrate {f(t).g(-(t - t_0) }. You need to flip g(t) to g(- t) and then shift it t_0 unit to get g(-(t-t_0)). Thus you need to transform [1 -1] to [-1 1] before shifting it to do the convolution.

In project 1, we normally use symmetric kernel like [1 2 1] for LPF and thus it doesn't matter whether we flip it or not. However, to do correct convolution, you definitely need to flip the g(t) before convolution.

Q2: The kernel coefficients must be summed up to 1 or you will increase intensity( >1) or decrease intensity ( < 1) of the pixel. Some of you forget to put 1/256 in front of the 5 by 5 matrix.

It seems that not too many people have signed up for demos. Please sign up soon. If you don't give a demo, we cannot grade you.

Please sign up for a Project 1 Demo slot. The signup sheet is outside of the door where Yu-Chi sits (1346 CS).

We really don't want to schedule more slots because people forgot to sign up.