How the foreground region is determined for downstream object detection tasks

[HenryHoou]

Thanks for your excellent work！
You mentioned in TPAMI paper that predicting the complete shape of objects in the foreground category goes in parallel with categorizing and determining the box, so are the details of this step going to isolate the foreground category and proceed to completion or are you going to directly input the larger region?What is your method for determining the foreground？