Skip to content
Permalink
Browse files
reverted to L2 regularization for splined models
The difference-from-mean regularization wasn't good as a "regularization". Any
place would be a perfect fit as far as the regularization is concerned. So the
optima found by the solver would end up with giant focal lengths, compensated
for by crazy intrinsics
  • Loading branch information
Dima Kogan authored and Dima Kogan committed Aug 23, 2020
1 parent 6039abe commit c8f9918023142d7ee463821661dc5bcc8f770b51
Showing with 40 additions and 338 deletions.
  1. +40 −338 mrcal.c

0 comments on commit c8f9918

Please sign in to comment.