MixFormer icon indicating copy to clipboard operation
MixFormer copied to clipboard

Test on our own dataset with multiple instances per video

Open samueleruffino99 opened this issue 1 year ago • 3 comments

Hello, I would like to adapat the datasets to handle multiple instacnes per video (instead of havinf only one object per video). Is it already handling this situation (performing SOT task multiple times per video on every single object)? Or is there a way to adapt the current loader? One way is also to generate multiple video instances per tracked object and treat them as separate video, but I think is not the straighter approach.

Thank you!

samueleruffino99 avatar Nov 12 '23 12:11 samueleruffino99

There is no implementation of such pipeline.

songtianhui avatar Nov 12 '23 13:11 songtianhui

I have seen you have a multiobj_mode param in external/AR/pytracker/tracker, is that referring to this mode or it is something different? Thank you!

samueleruffino99 avatar Nov 16 '23 09:11 samueleruffino99

sorry, the parameter is not used.

yutaocui avatar Nov 17 '23 13:11 yutaocui