Project2 / GroupJeffWillRyanBlogMay 9,2009 by zoerb (link) Here's a shot of the difference between two frames of two different motions (x-axis = motion1, y-axis = motion2). Black means no difference, white means high amount of difference. The algorithm has been in place for a while, but I just wanted to output its results as an image so we could better visualize it. May 7,2009 by mccardel (link) We've got 3 motions shown at once working fairly well, so it's now possible to actually see how the splicing went. May 7,2009 by mccardel (link) It's been a busy week for us all, so progress hasn't been nearly as steady as we would've liked. That said, as long as we don't hit any massive bumps, we'll be on track to finish getting a basic dynamic timewarping completed by the demo time. I've worked out a few kinks with the current version of our dynamic table, and am now moving on to fixing and adding the rest of the parts for dynamic timewarping. May 1,2009 by mccardel (link) Thanks to some stuff Ryan did, we now have basic splicing. Doesn't look bad either. We still need to get dynamic timewarping to work, but that's for next week! April 29,2009 by mccardel (link) Currently we're monkeying around with getting Dynamic Timewarping to work. Unfortunately, right now it's more "monkeying" and less "working" :( April 24,2009 by zoerb (link) We will be implementing motion splicing (as in number 4 on the project page) for our third project. For background and details on how to implement this we will be relying mostly on this paper along with this one for timewarping. Our current plan is to have a naive version of splicing working by early next week. This version will simply orient and attach an upper-body motion onto a lower-body motion. In the following week we will work on getting timewarping working, and the week after that will be used to put everything together, fix any bugs, etc. If we have any extra time, we might try doing something like the posture transfer as in the first paper. April 15,2009 by jhugo (link) i'm working on writing to a file. it'll be a function in the bvh_loader library that will take a bvh_file object April 15,2009 by mccardel (link) Concatenation works just fine (As far as I know at least!), and Ryan is currently working on blending. We still need to be able to write to files, find a better way to show motions, and get the camera in a good position. Unless all of our copies (and the svn server) are wiped, there's no reason we won't be done on friday. April 14,2009 by jhugo (link) Will is right in that the rotation information order supported in ZXY. The infrastructure is there to support other ordering, its just not fully implemented yet. The BVH parser is a bit picky. I'm rather sure that it will quit loading on error and not bring down the program, but it will leak memory. I've basically ran it on random bvh files looking for errors. It won't accept blank lines. It expects the heirarchy before the motion section. It'll only take one heriarchy section (I understand how two could possibly work, but it'd think that it would be two different files anyways. I haven't found a bvh file that has two heriarchies anyways). It does require a name after the joint or root. only 6 channels for the root and 3 for the joints are supported. I suspect there are not enough safe guards when parsing the actual motion data, and I haven't fully tested multiple motions in a file (having an example file would help) but everything seems ok. April 12,2009 by mccardel (link) I believe the only motion we have down (parser-wise) is the ZXY one. We should be able to get at least a few more done by the end of the project though. We're thinking about doing some sort of interpolation based on physics if we can find the time. April 10,2009 by zoerb (link) We're not quite sure which special features we want to do yet. April 3,2009 by jhugo (link) I have some code that will parse the tree part (header) of a bvh file and generate a tree. There are a few bugs i need to work out yet, so i'll likely commit sometime tomorrow. March 31,2009 by jhugo (link) svn repo is up, and you should have access the repo path is /u/j/h/jhugo/svnrepo/cs638_p2 below is a nice CSL tutorial for using svn on unix. I'd suggest tortoisesvn for windows http://www.cs.wisc.edu/twiki/bin/view/CSDocs/FileSharingAndSVNHowTo March 30,2009 by mccardel (link) Our current plans are to use C++ and fltk, and use something like boost for various data structures. March 26,2009 by gleicher (link) empty |
