In trackerGNN, why is the costOfNonAssignment half of the Assignment Threshold?

1 view (last 30 days)
In trackerGNN, why is the costOfNonAssignment half of the Assignment Threshold?

Accepted Answer

Elad Kivelevitch
Elad Kivelevitch on 17 Mar 2022
Because the penalty for not assigning a detection to a track is applied twice - once for the unassigned detection and once for the unassigned track.
  2 Comments
Ozlem Gumuskanat
Ozlem Gumuskanat on 18 Mar 2022
Thank you for your interest and time. Is this because of assignment process? I mean, is this penalty proceed in auction for example? Is this because matrix multiplication in matrix multiplication? Thank you in advance.
Elad Kivelevitch
Elad Kivelevitch on 29 Apr 2022
It has nothing to do with the choice of assignment algorithm or matrix multiplication. It is just part of the assignment cost. Let's say you have only one track and one detection with a cost of assignment being 20 and cost of non-assignment being 15.
So, the optimal solution is actually gained from assigning the detection to the track, because you incur a cost of 20. If you didn't assign them, the unassigned track would cost you 15 and the unassigned detection would cost you another 15, leading to a total cost of 30.
Because it could be confusing why a cost of 20 is better than a cost of nonassignment of 15 we decided to ask for the 30 and split it in half internally.

Sign in to comment.

More Answers (0)

Products


Release

R2020a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!