Your implementation of _cluster in lmb.py is to collect clusters of targets. I think this generator is used for Grouping and Gating as shown in the below image.

Is your generator based on the below image? where target groups are formed based on targets that share at least one measurement.

Hi,
I am using your code to track people using MOT17 dataset in my LMB respository.
I have tried for more than two weeks to find a reason why the label for the same target is overlapped (you can check the following images) but failed.

Multiple labels are generated for the same target.

If I increase r_lim from 0.0025 to 0.003 (Kill off targets with low probability). the label of each target is generated every new frame.