Train on KINS dataset

[McCGreen]

Excuse me!
There is a detail question trouble me.
In AISFormer/detectron2/data/datasets/coco.py. Line383~391
Here you have three variant 1.segm 2.bo_segm 3.i_segm. It seems that here you have three segmentation.
In COCOA, keys contain 'segmentation', 'visible_mask', 'invisible_mask'.
But in KINS's json file, the keys are 'segmentation', 'area', 'inmodal_seg', 'iscrowd', 'image_id', 'bbox', 'inmodal_bbox', 'category_id', 'id'.
You can see only segmentation and inmodal_seg(visible segmentation) is available.
How can I Corresponds to these three variables?