Package Summary

This package performs object pose refinement by using ICP and/or optimization over a beam-based sensor model. For scenarios with large amounts of clutter, occlusion, and pose uncertainty, this package offers functionality for performing random restarts with different evaluation functions, including a segmentation-aided beam-based sensor model.

Notes

When installing, it is highly recommended to have CUDA installed, though not strictly required. A message during compilation will indicate whether or not CUDA has been detected on your system. If detected, it will be automatically used for the beam-based sensor model error evaluation.

For test programs, including tools for running pose_corrector over bag data and a preliminary integration with tod, see pose_corrector_test.

To perform pose refinement as part of an Ecto graph, Python bindings for this package are available in ecto_corrector.

Wiki: pose_corrector (last edited 2011-09-19 04:13:33 by MichaelKrainin)