Is box regression method predict_boxes applicable to tracking multiple objects, number of interested objects>2

[Akola-Mbey-Denis]

Please I would like to know if the predict_boxes method can be used to regress bounding boxes in a tracking task were we are not just interested in one object of interest. For instance,people and cars in videos.
Please suggest what I can do to achieve this if it is possible. I used the current method ,but it fails when I have more than 2 classes at the resize_boxes line 44 in code file frcnn_fpn.py. I am using specifically KITTI Object detection dataset with 9 classes