cheind / py-motmetrics

:bar_chart: Benchmark multiple object trackers (MOT) in Python
MIT License
1.37k stars 258 forks source link

Testing on different classes #69

Closed tiberium24 closed 4 years ago

tiberium24 commented 4 years ago

Hello,

Thank you for your great work.

I would like to test on a car tracking dataset. Do I just change ground truth and detection files as: ground truth: frame objectID xmin ymin width height 1 -1 -1 -1 test results: frame objectID xmin ymin width height -1 -1 -1 -1

Does it make sense to run motmetrics.apps.eval_motchallenge on those files for cars?

cheind commented 4 years ago

yes, that seems reasonable to me. MOT layout is as follows 'FrameId', 'Id', 'X', 'Y', 'Width', 'Height', 'Confidence', 'ClassId', 'Visibility', 'unused'

Note that the parser assumes matlab convention where image locations start at (1,1) instead of (0,0).

tiberium24 commented 4 years ago

Thanks a lot! And for your quick response!