Why need CLIP-Aligned Labeling on top of Region Prompting?

[Jasonkks]

Dear Authors,

Thanks for the great work.

I understand that the proposed Region Prompting already adapts the CLIP model to match the ground truth classes for base class objects. I am not sure why there is still a need to relabel the classes with 'CLIP-Aligned Labeling'?

Please correct me if my understanding is wrong. Thank you!