Project2 / GroupBetsyYangMikBlogMay 7,2009 by mikola (link) I got sick of waiting for the motion graph construction to finish so I implemented gradient descent for finding the local minima. The result is that I can now construct a motion graph with over 8000 frames in less than 1 minute. Path following is still really slow. I tried improving performance by adding some constraints on the velocity with little effect. May 6,2009 by mikola (link) Got path following working. Checked in an example file to show it in action. The only problem: It is very slow. Motion graph synthesis takes a long time and second the actual path generation takes even longer. I have some ideas for how to make the motion graph generation more efficient (basically replace the brute force n^2 test with Newton's method.) For the path searching, I think that an A^* branch and bound test would definitely perform better than the current horrible brute force DFS I am using right now. I will work on tweaking the implementations if time permits. May 4,2009 by mikola (link) Here is the surface plot of the motion self-similarity for a walking clip:  April 15,2009 by elisabet (link)  Yeah, I'm using the old GUI, it's doing some weird stuff on the new one that I need to poke at more (it might just be that the camera's too far away so it's drawing too small) April 12,2009 by mikola (link) Here is a grocery list of features:
Now for the future stuff:
And here are some things that would be really neat, but probably won't get done due to the fact everyone is going to be busy as hell next week:
I'm going out of town next Friday so I'll have to skip out on the class presentation. Should be fine without me. April 10,2009 by yangk (link) Added basic projection shadow onto the ground plane, reflection, a new ground plane, a render option menu, and fixed the resizing of the opengl window to work properly. The makefile includes a few more library includes 'LNK_LIBS = -lglut -lm -lfltk_images -lpng -lz -ljpeg -lfltk -ldl -lXext -lX11 -lfltk -lfltk_gl'. I think there is an easier way to specify the fltk library includes by using a shell script provided in the fltk source which is something like 'fltk-config --use-images --libs --ldflags' which basically returns '-lfltk_images -lpng -lz -ljpeg -lfltk -ldl -lXext -lX11 -lfltk -lfltk_gl' but I'm not sure how to put that inside the make file. Let me know if any of you are having trouble compiling it. I got it run in both Windows and Linux.  April 8,2009 by yangk (link) I checked in the GUI code which uses FLTK. I haven't tested it fully for bugs and crashes, but I noticed it crashed a few times when using motion concatenation. You will need FLTK 1.1.9 which can be downloaded here http://www.fltk.org/ I got the code to compile under Windows and Linux. I checked in a Visual Studio 2003 project in case someone wants to work in Windows. In Windows, the glut library is also needed which can be downloaded here http://www.xmission.com/~nate/glut.html and the Eigen library as well. In Linux, I compiled the FLTK library but did not install them so in my makefile, I had to specify the path of the includes and library like the snippet below. I'm not too familiar with Linux so I wasn't sure if I could install the library on the lab computers.
FLTKPATH = /afs/cs.wisc.edu/u/y/a/yangk/Desktop/p2/fltk-1.1.9 ..........
INC_PATH = -I$(srcdir) -I$(EIGENPATH) -I$(FLTKPATH)
LNK_LIBS = -lglut -lm -L$(FLTKPATH)/lib -lfltk -lfltk_gl April 7,2009 by mikola (link) Time: Today at 4:30pm Location: Room 2054 in the Mechanical Engineering building, Spatial Automation Lab Need to figure out what is left to do, distribute responsibilities. April 6,2009 by mikola (link) Here's a quick screen grab of motion concatenation/interpolation in action:  April 5,2009 by mikola (link) The interpolation now works for quaternions. Used the logarithm implementation from Eigen2 to get correct shortest paths. Still haven't implemented the conversion from quats -> Euler angles or Euler angle interpolation. Next on the list is multiple motion concatenation, but probably best to wait on further modifications until Monday.  April 3,2009 by mikola (link) Decided to implement ellipsoid drawing as well as the quaternion parameterization (though one is more visually obvious than the other.) No shading yet, but adding that in is relatively trivial.  April 3,2009 by mikola (link) Email: mikolalysenko@gmail.com Schedule:
It looks like next Tuesday might be a good time to meet up, at around 4:30 or so. I need to figure out how to get the test data from AFS since my workstation is on the CAE network. April 2,2009 by yangk (link) My email is yangk@cs.wisc.edu Here's my schedule in case we want to schedule a meeting.
So basically, I'm available tuesday, thursday, friday nights and weekends. Some times might not work as well if I can't arrange a babysitter in advance. I will be out of town this weekend. April 2,2009 by mikola (link) Created a google code repository: http://code.google.com/p/skeletal-anim-838 Still need to get emails from Betsy and Yang so they can be added to the list of project administrators. Need to figure out how to split up the rest of the project. Here is a list of the libraries that are so far included in the project:
The code is written in C++ and currently builds via the GNU Makefile system. Data files for meshes are read from stdin. April 1,2009 by mikola (link) Got basic skeleton loading working, wrote a quick GLUT shell to draw a point at the origin of each joint's coordinate system.  March 26,2009 by gleicher (link) nothing meaningful yet |