symforce.examples.bundle_adjustment_fixed_size.generate_fixed_problem module#

class FixedBundleAdjustmentProblem(num_views, num_landmarks)[source]#

Bases: object

The setup is that we have N camera views for which we have poses that we want to refine. Camera 0 is taken as the source camera - we don’t optimize its pose and treat it as the source for all matches. We have feature correspondences from camera 0 into each other camera. We put a prior on the relative poses between successive views, and the inverse range of each landmark.

This is called from symforce/test/symforce_examples_bundle_adjustment_fixed_size_codegen_test.py to actually generate the problem

Parameters:
  • num_views (int) –

  • num_landmarks (int) –

generate(output_dir)[source]#

Generates functions from symbolic expressions

Parameters:

output_dir (str | PathLike) –

Return type:

None