Use MST initialization with huber loss#802
Use MST initialization with huber loss#802ayushbaid wants to merge 5 commits intofeature/shonan_mst_initfrom
Conversation
|
This seems to do better
|
ayushbaid
requested review from
akshay-krishnan,
dellaert,
johnwlambert and
travisdriver
|
@ayushbaid the diff is against "MST w/o Huber", not "no MST" right? Are we still seeing regressions when compared to "no MST"? |
The metrics are compared against master though |
Yes, I meant the diff in metrics. |
This commit passes in the optimality treshold explicitly from GTSFM: a value of `-1e-5` compared to `-1e-4` default value. I tried using 0 and `1e-5` (value used in paper for large datasets) but even the unit tests are taking up a lot of time with higher thresholds.
travisdriver
left a comment
travisdriver
left a comment
There was a problem hiding this comment.
Why did you turn off inlier weighting?
|
Also, are you sure inlier weighting was turned on it master when you compared? I vaguely remember @akshay-krishnan accidently turning it off in master. |
It was just for experimentation and comparing different sets of values. Inlier weighting is on right now on master. |
3 participants
No description provided.