[Work Log] Work Log
February 19, 2015
Kalman smoother is tested and appears to be working well. 70% faster than direct method.
Queue
- Investigate "zero covariance" bug
- infer liklihood variance during reconstruction
- build pipeline, w/ intermediate results
- project 3D posterior into third view
Later
- after projecting into third view: Add-remove curves grow/shrink curves ** add/break branches
Today: Cleaning up and building pipeline
- move logic from tmp_track_all to a function
- write pipeline script - all input/output in one directory
- write intermediate output at each stage curve extraction branch estimation warping likelihood estimate pair-wise 3d reconstruction 3rd-view warping likelihood estimate 3-view reconstruction ... nth view warping ** n-th view reconstruction
for each view, keep likelihood estimate
Posted by Kyle Simek