As mentioned, there is a project checkpoint due on Friday, March 11th. Because I haven’t announced the details until late, I will give you some flexibility as to when you turn it in (but definitely before Monday, March 14th)
Each group should make a posting to the “Project 1 Pre-Break Checkpoint” category. In this posting you should:
- Respond to any feedback that you have gotten (including the comment on each that I have added – although those may not be ready until later).
- Describe your progress since the original proposal. If you really haven’t done any work on the project, be honest about it – but get busy! Time will fly by.
- Describe any updates that you would make to your proposal.
If you have a picture of your system doing something, please include it.
Note that there will be much more expectation for the March 25th checkpoint.
{ 1 comment }
Jim:
****
Working on new framework and cleaning up the code and redesigning
interfaces.
Lulu:
*****
1. Completed per frame bounding box
2. Can track individual joints in the BVH file.
3. Can load multiple characters and play all of them in the same scene.
4. Presently working on selecting characters in the scene for editing purposes.
Raja:
*****
1. Completed coversion of keyframe data to Quaternions.
2. Completed rendering based on Quaternion keyframe data.
3. Working on interactive path creation.
Chaman
*********
Started with “Olga Sorkine’s Least Square Mesh paper” to implement the Laplacian Mesh Editing since it was much easy to understand than Igarashi’s
“As rigid as Possible Shape manipulation paper”.
First Raja wrote the Laplacian code which didn’t work and produced liner segments between the constrained nodes and later we realized that a straightforward Laplacian will always converge to straight lines only. So I started working on “4 Node Interpolatory Subdivision Curve Editing method”, which although produced smooth curves and in fact, too smooth everywhere which was unexpected from the curve editing point of view. So, I gave this method up. Then I carefully read the Olga’s paper, and implemented the least square laplacian code ( With Cholesky Solver for A’A x = A’b ) for sparse matrices ( Dense method took more than 10 minutes for 1000 nodes, so it was impractical). This method produces better results, but has the problems of shifting the entire curve and suffers from the non-local behaviour. There are some heuristics that I will try, but it seems that all simple Laplacian methods produced too smoothed curves everywhere, and none were close to Jehee Lee’s paper and video results.
So it seems that Igarshi paper is the only way to go, but first I need to understand its mathematical parts and implement it. This paper has 2 degree of freedom and we need to modify the method for 1 D curves, but from 2D to 1D although seems simpler, is not quite obvious to me at present (for 3D curves). Also I am extremely confused by the word “Laplacian” in the Lee’s paper, as the video shows that the paths between the constraints are straight lines and there is
no smoothness anywhere, so why and what Laplacian doing in their paper.
{ 1 trackback }