Skip to content

v1.0.0

  • v1.0.0
  • 37142d8
  • Verified

    This commit was created on GitHub.com and signed with GitHub’s verified signature.
  • Choose a tag to compare

  • v1.0.0
  • 37142d8
  • Choose a tag to compare

  • Verified

    This commit was created on GitHub.com and signed with GitHub’s verified signature.
@rlav440 rlav440 tagged this 02 May 02:23
* Added pytest as the testing library and fixed old test behaviours

* fixes for readme badges

* Fixes for small issues

* fix wrong axis direction in visible pose estimation

* fixed outlier removal, added adaptive decimation function

* added check for empty image folders to the abstract target

* added check for missing detection in target detections

* removed leftover fields from camera loader

* Moved functionality to class functions on the bundle primitive.

* removed a recursive callstack from targetDetection

* Added a fallback to non-compiled code for initialisation

* cleaned the requirements and dev setup

* updated the tests to pass

* setting up a tagged publishing workflow

* removed extra file

* testing thoughts to adjust to the paper input

* built a function block based definition of bundle adjustment

* fixed the jacobean logic

* compiled numba single jacobean line

* added caching to the compilation of the basic functions

* added working function eval

* sketched out function for computing loss from a detection array

* Added the supporting structure for running an optimisation

* rewriting the template handler to use the new function block based definition

* progress towards testing the new loss function

* fully running but incorrect values for just the loss

* optimised the code generation for customisable calibrations

* adding template tests to debug the loss function

* loss function works

* working loss, running jac

* testing_broken_jacobian

* working jacobian output

* starts to write the result caching logic

* ast syntax tree parser for determining token flow

* working multithreaded jacobean eval

* working loading of dynamic jacobean

* cleaning up the standard functions for calibration

* Working on the jac presentation

* Update setup.cfg

* further improvements to analytic jacobean

* removed extra file

* fixed the detection behaviour for the image detection when given empty data

* fixed a further bug

* updated version

* futher improving the analytic implementation

* fixed handling of empty data detections

* added visualisation of the calibration target points, and additional checks for calibration

* Added a test for self calibrating target, gave a unique id to each loss

* Passing tests with working self calibration based approaches. ready to merge!

* added the ability to do a self calibrating bundle adjustment

* Correct gauge mapping. WOOHOO

* fixed the linearity condition for gauge fixing

* added comments and fixed tests to use dynamic cpu counts

* rewrites code generation to use gathered params, static sparse jacobians

* updated the runtime graphs

* a compiled function implementing the multivariate calc

* accellerates calibration using compiled MVC flow

* added loggging warnings to AFB when generating a new function

* prepping

* added the free point handler

* Forced int64 casting for param vector indexing to prevent windows errors on compiled fuctions

* fixed nets for the ccube!

* removed prints from the target detection for ccubes

* Added a cropping mode to the cameras writing, allowing smaller faster reconstructions of rois

* improved the visualisation, especially for very large numbers of datapoints

* Improved benchmarking tool with time limit and improved rigid transform documentation

* removing old setup scripts

* changed target draw size from 1080,1920 to 480,640

* Improved the performance of camera undistortion and projection functions.

* Added a sane limit to the default number of threads, rather than all but 2

* removed duplicate ,

* Added the missing array undistort

* updated the visualisation to match the new output of triangulation

* prevented random choice from selecting cameras multiple times

* made non random ordering default for writing to text

* fixed numba compile error with forced types

* added functionality to find close cameras

* Updating the library to use the recent openCV charuco syntax

* removed need for contribs, as ARUCO is part of opencv now

---------

Co-authored-by: Robin <rlav44@aucklanduni.ac.nz>
Co-authored-by: raymon124 <theprolificone@hotmail.com>
Co-authored-by: Robin Laven <robinlaven@Robins-MacBook-Air.local>
Assets 2
Loading