Generalized intersection over union between axes-aligned cuboids
Our implementation works only for 3D case. The version for the cuboids aligned with Cartesian axes needs the detection vectors to have at least 6 variables.
In difference to the Mahalanobis metric where the variables are treated uniformly, we need to identify which variables represent position of the cuboids and which variables represent sizes of the cuboids. In order to define this, we use the keyword argument ind_pos_size of the method .set_association_metric()
...
tracker.set_association_metric('giou-axes-aligned', ind_pos_size=(0,1,2,3,4,5))
The keyword argument ind_pos_size defines indices of three position variables and three size variables. When the argument is omitted, the sequence is ind_pos_size=(0,1,2,-3,-2,-1). This choice encodes first variables in the measurement vector being position and last three variables corresponding dimensions
Such a choice fits the data layout of several datasets including the popular nuScenes dataset. See our suggestion for tracking nuScenes annotations. The performance of the axes-aligned GIoU is suboptimal comparing to GIoU for oriented cuboids.