Current State Of The Art In Instance Segmentation?

Hi. I'm a layman interested in using computer vision for instance segmentation. What's the current best way to get up and running with a(n open source) model that supports fine tuning or transfer learning? Are there any few shot models? I came across one called clipseg (https://github.com/timojl/clipseg) in my search.

art computervision segmentation state