Daniel Eaton Nonlinear optimization project 

I will complete my ICCV 2005 contest entry. I intend on (at least) matching the performance of the winning teams' entries. Back to the main page.

Progress:

  • Code. December 13, 2005.
  • Project Report. December 5, 2005. 15 pages.
    "Why you care?": I derive the exact equations to solve the structure and motion problem using bundle adjustment. They are a little tedious, so if you feel lazy, copy and paste away. I parameterize my cameras with the usual extrinsic matrices (rotation plus translation), but only account for the cameras' focal lengths in the intrinsic matrix K. If your application uses a richer camera model, then you'll have to do those derivatives yourself (though it should be fairly evident how to proceed, based on what I give you).
  • Presentation. December 5, 2005. 20 slides.
    Mandarin Orange Box Animation 1 (two-view). 6.4 MB.
    Mandarin Orange Box Animation 2 (three-view). 10 MB.
    World Map Animation (two-view). 1.1 MB.
    "Why you care?": The animations of the algorithm running on real data demonstrate what the static figures of the report cannot.
  • Literature review. November 25, 2005. 6 pages.
    "Why you care?": You don't! My derivations contain errors, see the report instead.
  • Project proposal. November 6, 2005. 1 page.
    "Why you care?": You don't.