Rn we have a test in tests/imaging/integ_test.py that runs on multiple frames that are annotated with camera pose plus 3D target position and then gives an average score out of 5 for how many targets we get within N meters (I think it's set to 2 or 3 rn). I want to use this to see how much positional noise we can have and still get a decent average on these tests (like 3/5). This would involve adding another test that throws in noise, and maybe generating more 3d datasets than the three we currently have, which is done by attaching this script: https://github.com/uci-uav-forge/godot-data-gen/blob/main/Scripts/gen_3d_testing_data.gd to the camera object in Godot and running it.
After completing this task, we can test out some fun ways to filter out the position noise like VSLAM with Kalman Filters or pose graphs.
Rn we have a test in
tests/imaging/integ_test.pythat runs on multiple frames that are annotated with camera pose plus 3D target position and then gives an average score out of 5 for how many targets we get within N meters (I think it's set to 2 or 3 rn). I want to use this to see how much positional noise we can have and still get a decent average on these tests (like 3/5). This would involve adding another test that throws in noise, and maybe generating more 3d datasets than the three we currently have, which is done by attaching this script: https://github.com/uci-uav-forge/godot-data-gen/blob/main/Scripts/gen_3d_testing_data.gd to the camera object in Godot and running it.After completing this task, we can test out some fun ways to filter out the position noise like VSLAM with Kalman Filters or pose graphs.