Skip to content

2013.04.03 H. Segmentation Meeting Report

Jump to bottom Edit New page
cotemyriam edited this page Apr 11, 2013 · 2 revisions

Sub Projects | Meeting Reports

Meeting with Mehdi

  • Mehdi shows initial Picassa results to Chris.
  • Chris suggests use of simple threshold to merge face boxes that are nearby in time and space and remove non-detect noise
  • Mehdi notes he is using some heuristics along these lines but will add some more
  • Chris notes that simple HMM formulation might be more principled way to do the merge
  • Chris notes his lab has state of the art face verifiers if needed but as initial cut a simple histogram comparison might get the job done
  • Chris notes the following tracking paper
  • http://www.sanjivk.com/faceTracking.pdf
  • Chris notes that longer term a CRF formulation might be good way to go
Add a custom footer
Add a custom sidebar
Clone this wiki locally