speed_direction_batch() at the beginning of tracking about oc_sort HOT 5 CLOSED

Thanks for the reply!
If it is possible, could you please update codes here (and in mmtracking). I used mmdetection for a while, and I think it makes codes harder to understand due to its modular design.

That's just my little wish, don't take it so serious.

from oc_sort.

I had a thought. When the previous_obs is [-1, -1, -1, -1, -1], the velocities is 0, which makes diff_angle_cos to be 0. So it doesn't matter.

Am I right?

from oc_sort.

Yes, you are right. The [-1,-1,-1,-1,-1] is just a dumb (and ugly) placeholder for the beginning when no historical observation is available yet, having no influence over the association matrix.

I apologize for the confusion. I should have made it a better way. Stay tuned for a better version here or in mmtracking. Or, please feel free to make PR for that!

from oc_sort.

Yes. I will definitely update this. The current released version is a timely and primary version. The coming version should be more complete and well-formed.

from oc_sort.

Thanks in advance! OC-SORT is truly awesome!

from oc_sort.