johnwlambert / argoverse_cbgs_kf_tracker Goto Github PK

Code to do detection of Car/Pedestrian using CBGS on Argoverse datasetis outsource ?

So,I ran the main script.
First I am actually working with the argoverse-tracking sample dataset.Will that be enough to run the code
Secondly,while running run_ab3dmot.py I ran into the following error:
cannot find folder argoverse-2020/validation/lidar/per_sweep_annotation_ammodal
From my understanding ,when we print the log ID it shoudl display a numeric value,but the output comes out to be lidar
Do you think this is because my pose_dir is set to argoverse-tracking-sample dataset.
Do I have to download the entire dataset
apologies for being pesky.

Hi,I was running the transforms_utils.py
During execution of the function test_yaw_to_quaternion3d(),it seems that the quaternion values computed by yaw_2_quaternion3d() have opposite signs (this starts at iteration 26)
For eg:
at iteration 27,Rotation.from_dcm(dcm).as_quat() has the following quaternion values
0.0 0.0 0.5183925683105252 -0.8551427630053461
whereas the function yaw_2_quaternion3d() prints the following values
0.0 0.0 -0.5183925683105252 0.855142763005346
Another example where the signs are switched for z and w
at iteration 41
Any ideas why the signs are being switched?I feel that the switching in signs may have a significant effect on tracking later on as

Hi, thanks for sharing the code. How can I know the detection performances?