JOURNAL




Week 1: June 1 - June 3

Back in my old stomping ground of Newell Simon Hall... This week I filled out all of the forms for computer logins and key card access. I spent the week working at home in my room. Yanxi was out of the office on Thursday and Friday, so we haven't come up with a master plan yet. I took pictures of textures on my trip to Italy which I added to Yanxi's master texture database. She wants to work on making it public this summer so that there is a large database of textures for researchers to use. I have been reading a lot of papers on texture synthesis and motion capture to get the background information that I need for the rest of the summer. While reading all of these papers I'm starting to realize that this isn't going to be an easy task. I have written a texture synthesis program before, but applying it to mocap data is going to be difficult. In one of the papers I read it mentioned that this could be an extension of their work, so I wonder if it has been tried. Yanxi told me that it hadn't, but it might be worth some investigation to see if there are results already out there. I also decided to start building this website as I had some free time. I had to re-learn HTML as I haven't used it since senior year of high school.

Week 2: June 6 - June 10

The beginning of the week was pretty good. I'm starting to get excited about the work I'm doing. I did a lot of research in my area and feel that I have learned more in the past week than I have in entire classes. Motion capture is a really interesting field. I found a paper by this girl at Harvard who's senior thesis is very similar to the work I'm going to be doing. It was encouraging to see that it can be done. The most frustrating thing in the world is to be held back by a lack of resources (at least for me). I wasn't able to be very productive until my computer was set up properly. On a happier note, the entire lab was invited to go to a Pittsburgh Pirates game on Tuesday, which was a load of fun (the Pirates even won!). Matlab is a great tool and pretty easy to learn and I'll be using it for all of my work. I've written one texture synthesis program already, and am now working on a couple of different variations which are more condusive to motion capture data. I decided to go to SIGGRAPH at the end of July which I'm really excited about. DMP is giving me funding which will help defray the costs.

Week 3: June 13 - June 17

This week I really got started with the project I'm doing. I met with Yanxi and James (the PhD student I'm working with) and we had a nice long pow-wow about the work I'm going to be doing this summer and what areas we want to concentrate on. After looking at my results that were similar in texture synthesis to Efros and Leung (1999) we thought we'd try methods developed by Efros and Freeman (2001). After I coded a program that found the best over lap for two motions and then greedily picked the best transition for each joint, I coded a program which used the industry standard of alpha blending the overlap. Unfortunately the blending looked better. James thought that we might get some better results and make the motions look less wobbly if we moved to quaternions. That's what I spent the end of the week looking at. Friday, I flew home to see my family. I haven't been there in 6 months. It will be good to get some home cooking. Here are some sample images of the motions that I'm using. The vertical axis is time and the horizontal axis is the number of joints. The first six joints are the translation and rotation of the root marker. The last few joints are the feet, and you can really see the periodicity there.

Marching


Singing in the Rain Jumps

Week 4: June 20 - June 24

All of the pieces of this crazy puzzle are starting to come together. I worked on a program which finds the Pearson's correlation coefficient between each of the joints. It returns a value between -1 and 1 depending upon if two series of numbers have a direct relationship or an inverse relationship, or no relationship at all. This will help us decide which joints must transition at the same time. I'm having quite a time trying to convert between Euler angles and quaternions. In all of the literature I have read on them it says "quaternions are great, but very unintuitive." No joke. It just about works...but not as well as I would like it to. I had a great weekend. It was really hot in Pittsburgh, so I went to the Schenely pool with a bunch of my friends. It felt good to go swimming and bake in the sun.

Week 5: June 27 - July 1

This is a short week for me as I'm going away on Wednesday afternoon for my annual trip to Fairfax, VA for the 4th of July. This trip will be extra special as my boyfriend Trey turns 21. He's throwing a huge party and friends are flying in from California just for the event. I'm so excited to see them all! We're going to King's Dominion Amusement Park on Thursday and going to see the giant fireworks display in D.C. on the actual 4th. I've never seen such a large display before. For the couple of days that I was at work, I fixed up my correlation code and made another attempt at fixing the quaternion code. The correlation code is now set, and the quaternion code has made progress, but I'm still not entirely happy with it. When I get back I will have a meeting with Yanxi and James to discuss where we want to go from here. On Tuesday we had a lab meeting and I got a preview of two SIGGRAPH talks that people from CMU are making. Its cool to see the behind the scenes preparation. I can't wait to go to the conference in a month. My funding came through from the DMP which I'm really happy about.

Week 6: July 4 - July 8

Going away was great. I got to see friends from school that I haven't seen since May. I even got a little bit of a tan. I actually met with Yanxi and James the morning before I left (the 29th). We talked about how we were going to use the correlation between joints in our synthesis. I've been writing programs to do this for the rest of the week. I'm finding it easier and easier to bang out the algorithms, but I'm getting really frustrated that nothing seems to be much of an improvement. I like to see results. Hopefully next week will be better. I need to figure outhow I can make movies out of my data in a format that can be put on the website. Lab meeting this week was cool because they were giving presentations on topics that I had a greater background in. The people in this lab are so darn smart. They seem to know everything about everything. Amazing. I'm glad that I have Vasu around because we can laugh about our lack of knowledge. When either of us is bored we can talk and remotivate ourselves. I have to fill out my mid-research report for the DMP this weekend. I feel I'm doing a lot of work, so it shouldn't be too hard. I'm looking forward to a weekend when I'm not going away.

Week 7: July 11 - 15

This week was again a shortened week as I went home on the 14th for my birthday weekend. I continued to work on all of my different algorithms and make sure they were free of bugs. I was able to speed up a few of them. I then took the marching motion and compared each of the algoritms on it. It takes quite a long time to run. I met with Yanxi on Wednesday and we talked about what we were going to do next. She wants to take her SIGGRAPH paper from 2004 and apply it to motions. We also talked about motion transplanting and verbs and adverbs and decided to read up on them and discuss at a later date.

Week 8: July 17 - 22

This week I focused on taking one motion and finding the 'average stride'. Yanxi's 2004 SIGGRAPH paper dealt with this idea in the texture realm. By finding the average stride of a motion and the different PCA bases, we will be able to vary the regularity of the motion. Calculating the mean stride took some time as I had to decide all of the cutoffs. I also got stuck with my dear friends the quaternions. It turned out that they weren't needed - thank goodness! Here is a link to to average cycle of a joyous walk. Hopefully I'll have time to make movies of the original cycles at some point, so you can see the difference.

Week 9: July 25 - 29

I completed all of the code for PCA this week. I found a great tutorial online which helped me understand exactly what I was doing why it was so useful. I spent a large amount of time breaking down motions into individual cycles. We decided that we didn't have enough cycles to calculate a meaningful mean cycle, so Yanxi booked a motion capture session where we were able to get the data that we needed. Yanxi also wants to look at sliding feet. She thinks that we will be able to detect sliding with our correlation code. Once I figure out a way to calculate the velocity of the feet, we can compare it to the velocity of the root. When the feet are sliding, the velocity of the root should be greater than the velocity of the feet. This should have a different correlation coefficient than regular walking. I also spent some time getting ready for SIGGRAPH, picking out which sessions that I want to attend. I'm really excited about going!

Week 10: August 1 - 5

This week was spent at the Los Angles Convention center at SIGGRAPH. It was a fabulous experience. I had never been to a conference before. Some things were as I expected them, but others weren't. The paper sessions were interesting. I got to see a few in areas that I'm working on now, mainly texture synthesis and styles of human motion. I was really impressed with myself that I understood the majority of the talks. Considering I only started studying graphics in January I have come a long way. I quickly learned that some sessions were more valuable than others. The courses hosted by animation studios spent a lot of time plugging their new movies and very little time talking about anything that I could use in my research. All of their software is proprietary so they can't divulge many details. I was really impressed with MASSIVE software. They participated in a course on crowd and group animation. They use PCA to vary strides of large groups of charactes so they don't look exactly the same. The conference was also good because I got to know the people in the lab better. We went out to eat several times over the course of the conference. Yanxi took me and her Ph.D students as well.

Week 11: August 8 - 12

Jessica asked all of the DMP students in the graphics lab to give 10 minute presentations at the weekly lab meeting next Tuesday. It will be good to see exactly what everyone else has been doing. I spent all of this week putting together my presentation. It was quite a lot of work. Making movies takes a long time because you want to get just the right angle and show off your results in the best way. James in is Japan so I have been sending the presentation to Yanxi for her review. After SIGGRAPH I know what good and bad powerpoint presentations look like. I hope that everything will go smoothly next week.

Week 12: August 15 - 19

The presentation went pretty well. It was really interesting to hear all of the feedback that I received. The biggest criticism was that we synthesized motion one degree of freedom at a time, instead of synthesizing the entire joint at once. To do this I would need to revert back to using quaternions. I'm hoping to find a player (maybe Maya?) that can play motion capture data when the angles are in quaternions. The results are really poor when I have used quaternions previously because the translation from quaternions to euler angles is not well defined. I met with Yanxi and one of the Ph.D. students from the lab to talk about this. We talked a lot about how we could correlate quaternions, which is basically a 4 element vector. It turns out that you can't correlate vectors, or at least no body really doesn't know how to do it. Yanxi decided that we should try and look at the angles between quaternions and try and correlate those.

Week 13: August 22 - 26

I can't believe that the summer is over already! All of the freshmen are on campus for their orientation week. The weather is so nice, it's hard to be indoors all day. I spent this week wrapping everything up and starting my paper. It is quite hard to write a technical paper - at least in a concise way! This summer has been a great experience. Other than something to put on a resume, it was a way for me to get a professor on a different level and work in an area that hasn't been explored before. I'm going to try my best to continue working with Yanxi next semester, but my schedule is pretty intense. We'll see how it goes. I really think that I learned more over the summer than I have in some entire semester classes. I hope that I'll find a job next year that allows me to do some graphics because it's something that I really enjoy doing.

Home About Me My Mentor Journal Final Paper