This repo exemplifies how to convert the label poses in KITTI tracking training split, to a usable rotation and translation that can be used to render vertically aligned bounding boxes that look like this

Sequence 00

Sequence 15

Sequence 20

This section of the code exemplifies how to get the pose with respect to any frame.