Track 1
Implement and compare 2-3 different trajectory matching algorithms
using:
-Dynamic Time Warping
-Hidden Markov Models
|
Track 2
This track is involving pose estimation. The program ideally would take images from
2+ cameras and would compare them to those generated for the 3D model.
The virtual world would generate
poses and attributes to match the real world, assigning fitness values as it goes, until
a pose is determined by the virtual world to match the real world. The project would
start out frame by frame with comparisons to the virtual world. It would then
generate images to match those occurring in the real world. However, after
an optimal algorithm is perfected the hope is that the program would run in real time,
in comparisons and image generation.
|
|
Goals:
My goals depend on what track I take. There are two possible
tracks that I will take for this independent study, dependent on what opportunities
look best after the code is cleaned up next semester. Both tracks would have a
goal of creating, and possibly publishing, a final paper.
|
Timeline:
Jan 17th - 26th:
Clean Up Code
Gather Links/Papers
Jan 27th - Feb 9th:
Document Code
Gather Links/Paper
Feb 10th - May 20th:
Track 1 or Track 2
More details
in the future.
|
Grading Criteria:
50% Project Progress and
Weekly documentation
25% Code Documentation
-commenting existing code -map of existing code
25% Final Paper
|