Is there any difference in the IDSW metric between TrackEval and motmetrics?

[mesllo-bc]

Hi, I'm in the process of adapting HOTA for analyzing a tracker. I've always used the py-motmetrics for this but I'm now also comparing TrackEval. I noticed that the number of ID switches and fragmentations I get there is slightly different compared to the number of switches I get here. Here it seems that I get slightly more (2 or 3 more counts). I would like to understand whether I'm doing something wrong or if this is just because the way the switches and fragmentations are counted differ slightly between the two repos, and if so, if you can kindly shed some light on where I can find more info about this. Thank you!