Open sfmig opened 2 weeks ago
Could 1-based indexing be slightly confusing in a Python environment? Could you end up with a situation where the 0th individual is named id_1
?
That's a fair point Adam. So we'd have to choose between a Python convention vs a MOT convention 🤷🏼
I slightly lean towards the Python convention, just to minimise the possibility of off-by-one errors creeping in, so id_0
, id_1
, etc.?
I also lean that way. If for no other reason than 100% of users will use Python, and some percentage <100 will be familiar with MOT.
Is your feature request related to a problem? Please describe. For the poses dataset, if no individual names are passed we name them starting from
individual_0
.For the bboxes dataset, we start with
id_1
(following 1-based convention from MOT challenge).With @niksirbi we discussed both should be consistent, and that
id_1
could be nicer because it's less verbose. So we suggest to change to 1-based ID naming for pose data.Describe the solution you'd like If no individual names are passed by default, name them as
id_1
,id_2
...Describe alternatives you've considered We could use 0-based indexing, but 1-based seems a lightly held standard at least in object tracking (MOT challenge).
Additional context \