How to further segment a subject?

Traditional instance segmentation is designed to distinguish different object instances and generate a binary mask for each instance. However, when dealing with objects like human bodies, one might need to further segment the instance into finer parts (e.g., head, torso, arms, legs), is there any way to acheive this goal?

Start by reviewing the repository's inference code and example notebooks to determine whether finer human-part segmentation is supported or would require a new approach. Done would need a clearly documented, reproducible path—or a scoped implementation—for producing separate masks for parts such as the head, torso, arms, and legs.

Written by the indexing model from the issue text.