Skip to content

Conversation

@pupueys
Copy link
Contributor

@pupueys pupueys commented Dec 17, 2025

  • actually fixed bug in yolo tracking and added full rgbd integraion. Made ros launch Node much clearer but adding a custom DynosamNode which is launchable. Testing co-planar issues with object tracking
  • testing filtering over motion in frontend
  • testing filtering but something broken in regular frontend motion
  • issue seems to be coming from using the pre-calculated of
  • working basic SRIF filtering estimating for frame to frame motion
  • testing SRIF with hybrid motion formulation and linearized point as parameter. Trick is to reset filter when we resample!
  • minor clean up
  • changed hybrid factors to finally use analytical jacobians and added a stereo hybrid factor for future use. Added unit tests on output (i.e residual) and Jacobian calculations
  • testing in prep for adding new visualisation for hybrid in both frontend and backend
  • added intermediate accessor extension of AccessorT. Made HyrbridAccesorCommon to bridge both hybrid methods and added keyframe visualisations in the backend display modules
  • publishing frontend-filtering keyframes for debugging
  • testing static Keyframing and utilising hybrid motion filtering estimation in backend
  • testing KF in backend with Hybrid and kF motion from frontend
  • Testing
  • testing docker build on jetson and refining relevant bash scripts
  • not compiling MultiSync due to jazzy vs. kilted differences in message filters
  • code compiling on ORIN
  • running NN on ORIN
  • fixed but in propogate dynamic obejcts and added architecture agnostic include file overwrites in cmakelists for openmpi
  • publishing frontend-filtering keyframes for debugging
  • bring in relevant things from filtetring branch but now VIODE is not working
  • Revert "running NN on ORIN"
  • solved conditional compilation of MultiSync
  • working version on desktop/orin (docker file permitting). Version is up to date with jm/filtering which has many unncessary thigns to run with current stable algorithms. Will backtrack KF hacks in front/backend before testing and then merging to main
  • intermediate readme that is under review
  • added option to turn off opencv display
  • removed old logging comments and updated new readme
  • improving timing stuff. Last commit before moving to new gtsam version
  • Update INSTALL.md with MPI error fix and dependencies
  • Update dataset download instructions in README
  • lies not moving to new gtsam
  • Revise MPI missing error section in INSTALL.md
  • Improve dataset download instructions in README
  • fixed relative rotation
  • updated official readme
  • added shebang to craete container files
  • fixed minor issues with dockerfile. Updated readme with more dynosam_nn related information and removed 'frontend' prefix from ground truth topics
  • testing for processing speed
  • testing timing in object detector
  • Added velocities to odometry message
  • Smaller fixes

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.

1 participant