1 paper · 1 filter
Xiaoke Huang, Jianfeng Wang, Yansong Tang +5
We propose a method to efficiently equip the Segment Anything Model (SAM) with the ability to generate regional captions. SAM presents strong generalizability to segment anything w…