Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Chordal initialization in Shonan #822

Merged
merged 7 commits into from
Feb 15, 2025
Merged

Chordal initialization in Shonan #822

merged 7 commits into from
Feb 15, 2025

Conversation

akshay-krishnan
Copy link
Collaborator

No description provided.

Copy link
Member

@dellaert dellaert left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Awesome! Maybe paste performance increase in comments.
But still some things to figure out:

  • Karcher or Anchor? Probably anchor (first key if not given)
  • Huber, certify, pMax
  • anchors from ground truth
    And see if that fixes the weirdness in those datasets.

@@ -52,14 +57,17 @@ def __init__(
super().__init__()
self._two_view_rotation_sigma = two_view_rotation_sigma
self._p_min = 3
self._p_max = 64
self._p_max = 3
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

64 is excessive, but 3 is maybe a mistake :-) We should monitor the number on a few datasets and choose based on that.

shonan_params = ShonanAveragingParameters3(lm_params)
shonan_params.setUseHuber(False)
shonan_params.setCertifyOptimality(True)
shonan_params.setUseHuber(True)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These things should not change, right?

Initial values as a gtsam.Values object.
"""
graph = gtsam.NonlinearFactorGraph()
anchor_key = None
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, we should figure out anchors in ground truth and use those if available.

@dellaert
Copy link
Member

dellaert commented Feb 5, 2025

Did latest changes fix those datasets?

@dellaert
Copy link
Member

@akshay-krishnan can we merge?

@akshay-krishnan
Copy link
Collaborator Author

Yeah, the unit test failure also happens on master, I'll fix that in a different PR

@akshay-krishnan
Copy link
Collaborator Author

Actually, I should first merge the TA PR

@akshay-krishnan akshay-krishnan merged commit e96796f into master Feb 15, 2025
23 of 24 checks passed
@akshay-krishnan akshay-krishnan deleted the sa-tweaks branch February 15, 2025 17:40
@dellaert
Copy link
Member

Yay !!!!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants