vlarsson
camera model support
Is it planned to support those camera models in near future?
FOV, THIN_PRISM_FISHEYE, RAD_TAN_THIN_PRISM_FISHEYE
You can checkout the dev branch. There are more camera models implemented including optimizers for their extra params. If you need to add a camera model you can just implement it into PoseLib/misc/camera_models.h and PoseLib/misc/camera_models.cc. If you implement the jacobian the optimizers should work.
If some model is still missing, see also here, which has Jacobians for most camera models: https://github.com/mapillary/OpenSfM/blob/d1165f9d522aee41f825be464a7e232286f529f1/opensfm/src/geometry/camera_distortions_functions.h. It has a permissive license (BSD-2), so you can look a the code and learn.
I would love to have RAD_TAN_THIN_PRISM_FISHEYE support in this library and be able to run PnP, and the link above has it. Porting it is above my (math skill level). This would allow me to work with the Meta Project Aria headset, whose camera is calibrated using this model.
You can checkout the
devbranch. There are more camera models implemented including optimizers for their extra params. If you need to add a camera model you can just implement it intoPoseLib/misc/camera_models.handPoseLib/misc/camera_models.cc. If you implement the jacobian the optimizers should work.
Coming back late...
If i checkout the dev branch, will it work with Glomap? Any idea?
@kocurvik
If i checkout the dev branch, will it work with Glomap? Any idea?
You would probably need to adapt it a bit. The dev branch uses a different way of passing options to the estimators.
Thanks for your quick answer. Puh, I dont think I am skilled enough to understand what that means.